1 tbd Sun Microsystems, Inc.
5 2.6.9-67.0.22.EL (RHEL 4),
6 2.6.16.60-0.31 (SLES 10),
7 2.6.18-92.1.22.el5 (RHEL 5),
8 2.6.22.14 vanilla (kernel.org)
9 * Client support for unpatched kernels:
10 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
11 2.6.16 - 2.6.22 vanilla (kernel.org)
12 * Client support for unpatched kernels:
13 we do not recommend using patchless RHEL4 clients with kernels
14 prior to 2.6.9-55EL (RHEL4U5).
15 * Recommended e2fsprogs version: 1.40.11-sun1
16 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
17 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
18 removed cwd "./" (refer to Bugzilla 14399).
19 * A new quota file format has been introduced in 1.6.5.
20 The format conversion from prior releases is handled transparently,
21 but releases older than 1.4.12/1.6.5 don't understand this new
22 format. The automatic format conversion can be avoided by running
23 the following command on the MDS:
24 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
25 For more information, please refer to bugzilla 13904.
26 * Output of lfs quota has been made less detailed by default,
27 old (verbose) output can be obtained by using -v option.
28 * File join has been disabled in this release, refer to Bugzilla 16929.
29 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
30 * NFS export disabled when stack size < 8192. Since the NFSv4 export
31 of Lustre filesystem with 4K stack may cause a stack overflow. For
32 more information, please refer to bugzilla 17630.
34 Severity : enhancement
36 Description: Update to RHEL5 kernel-2.6.18-92.1.22.el5.
39 Frequency : start MDS on uncleanly shutdowned MDS device
41 Descriptoin: aborting recovery hang on MDS
42 Details : don't throttle destroy RPCs for the MDT.
45 Frequency : race on file read and write
47 Description: Lustre doesn't delete files
48 Details : Clients drop lock reference and release openhandle when they find
52 -------------------------------------------------------------------------------
53 2008-12-31 Sun Microsystems, Inc.
55 * Support for kernels:
57 2.6.9-67.0.22.EL (RHEL 4),
58 2.6.16.60-0.31 (SLES 10),
59 2.6.18-92.1.17.el5 (RHEL 5),
60 2.6.22.14 vanilla (kernel.org)
61 * Client support for unpatched kernels:
62 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
63 2.6.16 - 2.6.22 vanilla (kernel.org)
64 * Client support for unpatched kernels:
65 we do not recommend using patchless RHEL4 clients with kernels
66 prior to 2.6.9-55EL (RHEL4U5).
67 * Recommended e2fsprogs version: 1.40.11-sun1
68 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
69 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
70 removed cwd "./" (refer to Bugzilla 14399).
71 * A new quota file format has been introduced in 1.6.5.
72 The format conversion from prior releases is handled transparently,
73 but releases older than 1.4.12/1.6.5 don't understand this new
74 format. The automatic format conversion can be avoided by running
75 the following command on the MDS:
76 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
77 For more information, please refer to bugzilla 13904.
78 * A new quota file format was introduced in 1.6.6/1.8.0.
79 The format conversion from prior releases is handled transparently,
80 but releases older than 1.6.6/1.8.0 don't understand this new
81 format. The automatic format conversion can be avoided by running
82 the following commands on the MDS and OSS servers (for
83 pre 1.4.12-1.6.5 quota files):
84 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV',
85 'tunefs.lustre --param="ost.quota_type=ug1" $MDTDEV'
86 or (for 1.4.12/1.6.5 quota files)
87 'tunefs.lustre --param="mdt.quota_type=ug2" $MDTDEV',
88 'tunefs.lustre --param="ost.quota_type=ug2" $MDTDEV'
89 For more information, please refer to bugzilla 13904.
90 * Output of lfs quota has been made less detailed by default,
91 old (verbose) output can be obtained by using -v option.
92 * File join has been disabled in this release, refer to Bugzilla 16929.
93 * A new Lustre ADIO driver is available for MPICH2-1.0.7.
94 * NFS export disabled when stack size < 8192. Since the NFSv4 export of
95 Lustre filesystem with 4K stack may cause a stack overflow. For more
96 information, please refer to bugzilla 17630.
99 Frequency : on remount
101 Description: external journal device not working after the remount
102 Details : clear dev_rdonly flag for external journal devices in
108 Description: shutdown vs evict race
109 Details : client_disconnect_export vs connect request race.
110 if client will evicted at this time - we start invalidate
111 thread without referece to import and import can be freed
117 Description: shrink LOV EAs before replying
118 Details : correctly adjust LOV EA buffer for reply.
123 Description: don't skip ost target if they assigned to file
124 Details : Drop slow OSCs if we can, but not for requested start idx.
125 This means "if OSC is slow and it is not the requested
126 start OST, then it can be skipped, otherwise skip it only
127 if it is inactive/recovering/out-of-space.
129 Severity : enhancement
131 Description: Update to RHEL5 kernel-2.6.18-92.1.17.el5.
134 Frequency : rare, need acl's on inode.
136 Description: client can't handle ost additional correctly
137 Details : if ost was added after client connected to mds client can have
138 hit lnet_try_match_md ... to big messages to wide striped files.
139 in this case need teach client to handle config events about add
140 lov target and update client max ea size at that event.
142 Severity : enhancement
144 Description: Update to sles9 kernel-2.6.5-7.314.
146 Severity : enhancement
148 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.31.
151 Frequency : Create a symlink file with a very long name
153 Description: ldlm_cancel_pack()) ASSERTION(max >= dlm->lock_count + count)
154 Details : If there is no extra space in the request for early cancels,
155 ldlm_req_handles_avail() returns 0 instead of a negative value.
160 Description: mds is deadlocked
161 Details : in rare cases, inode in catalog can have i_no less than have parent
162 i_no, this produce wrong order for locking during open, and
163 parallel unlink can be lock open. this need teach mds_open to grab
164 locks in resource id order, not at parent -> child order.
166 Severity : enhancement
168 Description: Add /proc entry for import status
169 Details : The mdc, osc, and mgc import directories now have
170 an import directory that contains useful import data for debugging
173 Severity : enhancement
175 Description: Re-disable certain /proc logging
176 Details : Enable and disable client's offset_stats, extents_stats and
177 extents_stats_per_process stats logging on the fly.
180 Frequency : Only on FC kernels 2.6.22+
182 Description: oops in statahead
183 Details : Do not drop reference count for the dentry from VFS when lookup,
184 VFS will do that by itself.
186 Severity : enhancement
188 Description: Generic /proc file permissions
189 Details : Set /Proc file permissions in a more generic way to enable non-
190 root users operate on some /proc files.
194 Description: Hitting mdc_commit_close() ASSERTION
195 Details : Properly handle request reference release in
196 ll_release_openhandle().
200 Frequency : only patchless client
201 Description: add workaround for race between add/remove dentry from hash
203 Severity : enhancement
205 Description: Allow OST glimpses to return PW locks
209 Description: LBUG when llog conf file is full
210 Details : When llog bitmap is full, ENOSPC should be returned for plain
215 Description: Prevent import from entering FULL state when server in recovery
219 Description: service mount cannot take device name with ":"
220 Details : Only when device name contains ":/" will mount treat it as
226 Description: replace ptlrpcd with the statahead thread to interpret the async
227 statahead RPC callback
231 Frequency : on recovery
232 Description: I/O failures after umount during fail back
233 Details : if client reconnected to restarted server we need join to recovery
234 instead of find server handler is changed and process self eviction
235 with cancel all locks.
237 Severity : enhancement
239 Description: Update to RHEL5 kernel-2.6.18-92.1.10.el5.
243 Description: Kernel BUG tries to release flock
244 Details : Lustre does not destroy flock lock before last reference goes
245 away. So always drop flock locks when client is evicted and
246 perform unlock regardless of successfulness of speaking to MDS.
248 Severity : enhancement
250 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.27.
252 Severity : enhancement
254 Description: Upcall on Lustre log has been dumped
255 Details : Allow for a user mode script to be called once a Lustre log has
256 been dumped. It passes the filename of the dumped log to the
257 script, the location of the script can be specified via
258 /proc/sys/lnet/debug_log_upcall.
263 Description: avoid messages about idr_remove called for id that is not allocated
264 Details : Move assigment s_dev for clustered nfs to end of initialization,
265 for avoid problem with error handling.
270 Description: avoid Already found the key in hash [CONN_UNUSED_HASH] messages
271 Details : When connection is reused this not moved from CONN_UNUSED_HASH into
272 CONN_USED_HASH and this prodice warning when put connection again
278 Description: avoid ASSERTION(client_stat->nid_exp_ref_count == 0) failed
279 Details : release reference to stats when client disconnected, not
280 when export destroyed for avoid races when client destroyed
281 after main ost export.
285 Description: more cleanup in mds_lov
286 Details : add workaround for get valid ost count for avoid warnings about
287 drop too big messages, not init llog cat under semphore which
288 can be blocked on reconnect and break normal replay, fix access
291 Severity : enhancement
293 Description: File striping can now be set to use an arbitrary pool of OSTs.
295 Severity : enhancement
297 Description: Export bytes_read/bytes_write count on OSC/OST.
301 Description: Early reply size mismatch, MGC loses connection
302 Details : Apply the MGS_CONNECT_SUPPORTED mask at reconnect time so
303 the connect flags are properly negotiated.
307 Description: Properly propagate oinfo flags from lov to osc for statfs
308 Details : restore missing copy oi_flags to lov requests.
312 Description: exports in /proc are broken
313 Details : recreate /proc entries for clients when they reconnect.
315 Severity : enhancement
317 Description: Add man pages for llobdstat(8), llstat(8), plot-llstat(8),
318 : l_getgroups(8), lst(8), routerstat(8)
319 Details : included man pages for llobdstat(8), llstat(8),
320 : plot-llstat(8), l_getgroups(8), lst(8), routerstat(8)
322 Severity : enhancement
324 Description: Implement lustre ll_show_options method.
326 Severity : enhancement
328 Description: Update to SLES9 kernel-2.6.5-7.312.
330 Severity : enhancement
332 Description: Update to RHEL4 kernel-2.6.9-67.0.22.EL.
336 Description: exports in /proc are broken
337 Details : recreate /proc entries for clients when they reconnect.
341 Description: don't fail open with -ERANGE
342 Details : if client connected until mds will be know about real ost count
343 get LOV EA can be fail because mds not allocate enougth buffer
348 Description: Resolve device initialization race
349 Details : Prevent proc handler from accessing devices added to the
350 obd_devs array but yet be intialized.
352 Severity : enhancement
354 Description: configure's --enable-quota should check the
355 : kernel .config for CONFIG_QUOTA
356 Details : configure is terminated if --enable-quota is passed but
357 : no quota support is in kernel
359 Severity : enhancement
361 Description: Update to SLES10 SP2 kernel-2.6.16.60-0.23.
363 Severity : enhancement
365 Description: Update to RHEL5 kernel-2.6.18-92.1.6.el5.
369 Frequency : rare, on PPC clients
370 Description: don't swab ost objects in response about directory, because
372 Details : bug similar bug 14856, but in different function.
374 Severity : enhancement
376 Description: lfs quota tool enhancement
377 Details : added units specifiers support for setquota, default to
378 current uid/gid for quota report, short quota stats by
379 default, nonpositional parameters for setquota, added
380 llapi_quotactl manual page.
382 Severity : enhancement
384 Description: *optional* service tags registration
385 Details : if the "service tags" package is installed on a Lustre node
386 When the filesystem is mounted, a local-node service tag will
387 be created. See http://inventory.sun.com/ for more information
388 about the Service Tags asset management system.
392 Description: Client runs out of low memory
393 Details : Consider only lowmem when counting initial number of llap pages
396 Description: Kernel BUG tries to release flock
397 Details : Lustre does not destroy flock lock before last reference goes
398 away. So always drop flock locks when client is evicted and
399 perform unlock regardless of successfulness of speaking to MDS.
402 Frequency : occasional
404 Description: add refcount for osc callbacks, so avoid panic on shutdown
406 Severity : enhancement
408 Description: Update to RHEL4 kernel-2.6.9-67.0.20.
411 Frequency : testing only
413 Description: sanity test 65a fails if stripecount of -1 is set
414 Details : handle -1 striping on filesystem in ll_dirstripe_verify
417 Frequency : only in unusual configurations
419 Description: Kernel panic with find ost index.
420 Details : lov_obd have panic if some OST's have sparse indexes.
422 Severity : enhancement
424 Description: Update to RHEL5 kernel-2.6.18-53.1.21.el5.
427 Frequency : rarely, if filesystem is mounted with -o flock
429 Description: do not process already freed flock
430 Details : flock can possibly be freed by another thread before it reaches
431 to ldlm_flock_completion_ast.
434 Frequency : rarely, if filesystem is mounted with -o flock
436 Description: LBUG during stress test
437 Details : Need properly lock accesses the flock deadlock detection list.
440 Frequency : rarely, if binaries are being run from Lustre
442 Description: oops in page fault handler
443 Details : kernel page fault handler can return two special 'pages' in
444 error case, don't try dereference NOPAGE_SIGBUS and NOPAGE_OMM.
447 Frequency : rarely, during shutdown
449 Description: timeout with invalidate import.
450 Details : ptlrpcd_check call obd_zombie_impexp_cull and wait request which
451 should be handled by ptlrpcd. This produce long age waiting and
452 -ETIMEOUT ptlrpc_invalidate_import and as result LASSERT.
458 Description: ASSERTION(CheckWriteback(page,cmd)) failed
459 Details : badly clear PG_Writeback bit in ll_ap_completion can produce false
463 Frequency : only with broken builds/installations
465 Description: no LBUG if lquota.ko and fsfilt_ldiskfs.ko are different versions
466 Details : just return an error to a user, put a console error message
468 Severity : enhancement
470 Description: Update to RHEL5 kernel-2.6.18-53.1.19.el5.
472 Severity : enhancement
474 Description: Update to RHEL4 kernel-2.6.9-67.0.15.
476 Severity : enhancement
478 Description: enable MGS and MDT services start separately
479 Details : add a 'nomgs' option in mount.lustre to enable start a MDT with
480 a co-located MGS without starting the MGS, which is a complement
481 to 'nosvc' mount option.
484 Frequency : always, on big-endian systems
486 Description: cleanup in ptlrpc code, related to PPC platform
487 Details : store magic in native order avoid panic's in recovery on PPC
488 node and forbid from this error in future. Also fix posibily
489 of twice swab data. Fix get lov striping to userpace.
492 Frequency : rarely, if replay get lost on server
494 Description: server incorrectly drop resent replays lead to recovery failure.
495 Details : do not drop replay according to msg flags, instead we check the
496 per-export recovery request queue for duplication of transno.
499 Frequency : after recovery
501 Description: precreate to many object's after del orphan.
502 Details : del orphan st in oscc last_id == next_id and this triger growing
503 count of precreated objects. Set flag LOW to skip increase count
504 of precreated objects.
507 Frequency : rare, on clear nid stats
509 Description: ASSERTION(client_stat->nid_exp_ref_count == 0)
510 Details : when clean nid stats sometimes try destroy live entry,
511 and this produce panic in free.
514 Frequency : occasionally since 1.6.4
516 Description: Stack overflow during MDS log replay
517 Details : ease stack pressure by using a thread dealing llog_process.
520 Frequency : very rare
522 Description: MDT cannot be unmounted, reporting "Mount still busy"
523 Details : Mountpoint references were being leaked during open reply
524 reconstruction after an MDS restart. Drop mountpoint reference
525 in reconstruct_open() and free dentry reference also.
530 Description: wait until IO finished before start new when do lock cancel.
531 Details : VM protocol want old IO finished before start new, in this case
532 need wait until PG_writeback is cleared until check dirty flag
533 and call writepages in lock cancel callback.
538 Description: mds_mfd_close() ASSERTION(rc == 0)
539 Details : In mds_mfd_close(), we need protect inode's writecount change
540 within its orphan write semaphore to prevent possible races.
543 Frequency : rare, on shutdown ost
545 Description: don't hit live lock with umount ost.
546 Details : shrink_dcache_parent can be in long loop with destroy dentries,
547 use shrink_dcache_sb instead.
550 Frequency : only when echo_client is used
552 Description: don't panic with use echo_client
553 Details : echo client pass NULL as client nid pointer and this produce NULL
557 Frequency : Always on 32-bit PowerPC systems
559 Description: fix build on PPC32
560 Details : compile code with -m64 flag produce wrong object file for PPC32.
565 Description: MDS LBUG: ASSERTION(!IS_ERR(dchild))
566 Details : In reconstruct_* functions, LASSERTs on both the data supplied
567 by a client, and the data on disk are dangerous and incorrect.
568 Change them with client eviction.
570 Severity : enhancement
572 Description: skiplist implementation simplification
573 Details : skiplists are used to group compatible locks on granted list
574 that was implemented as tracking first and last lock of each lock
575 group the patch changes that to using doubly linked lists
579 Description: delete compatibility for 32bit qdata
580 Details : as planned, when lustre is beyond b1_8, lquota won't support 32bit
581 qunit. That means servers of b1_4 and servers of b1_8 can't be
582 used together if users want to use quota.
585 Frequency : only with administrator action
587 Description: mount failure if config log has invalid conf_param setting
588 Details : If administrator specified an incorrect configuration parameter
589 with "lctl conf_param" this would cause an error during future
590 client mounts. Instead, ignore the bad configuration parameter.
593 Frequency : blocks per group < blocksize*8 and uninit_groups is enabled
595 Description: ldiskfs error: XXX blocks in bitmap, YYY in gd
596 Details : If blocks per group is less than blocksize*8, set rest of the
600 Frequency : Application do stride read on lustre
602 Description: The read performance will drop a lot if the application does
604 Details : Because the stride_start_offset are missing in stride read-ahead,
605 it will cause clients read a lot of unused pages in read-ahead,
606 then the read-performance drops.
610 Description: more ldlm soft lockups
611 Details : In ldlm_resource_add_lock(), call to ldlm_resource_dump()
612 starve other threads from the resource lock for a long time in
613 case of long waiting queue, so change the debug level from
614 D_OTHER to the less frequently used D_INFO.
616 Severity : enhancement
618 Description: add -gid, -group, -uid, -user options to lfs find
620 Severity : enhancement
622 Description: ll_recover_lost_found_objs - recover objects in lost+found
623 Details : OST corruption and subsequent e2fsck can leave objects in the
624 lost+found directory. Using the "ll_recover_lost_found_objs"
625 tool, these objects can be retrieved and data can be salvaged
626 by using the object ID saved in the fid EA on each object.
631 Description: this bug _only_ happens when inode quota limitation is very low
632 (less than 12), so that inode quota unit is 1 at initialization.
633 Details : if remaining quota equates 1, it is a sign to demonstate that quota
634 is effective now. So least quota qunit should be 2.
638 Description: Hung threads in invalidate_inode_pages2_range
639 Details : The direct IO path doesn't call check_rpcs to submit a new RPC once
640 one is completed. As a result, some RPCs are stuck in the queue
645 Description: Procfs and llog threads access destoryed import sometimes.
646 Details : Sync the import destoryed process with procfs and llog threads by
647 the import refcount and semaphore.
651 Description: mds fails to respond, threads stuck in ldlm_completion_ast
652 Details : Sort source/child resource pair after updating child resource.
657 Description: kernel BUG at ldiskfs2_ext_new_extent_cb
658 Details : If insertion of an extent fails, then discard the inode
659 preallocation and free data blocks else it can lead to duplicate
664 Description: don't always update ctime in ext3_xattr_set_handle()
665 Details : Current xattr code updates inode ctime in ext3_xattr_set_handle()
666 In some cases the ctime should not be updated, for example for
667 2.0->1.8 compatibility it is necessary to delete an xattr and it
668 should not update the ctime.
672 Description: add quota statistics
673 Details : 1. sort out quota proc entries and proc code.
674 2. add quota statistics
679 Description: quotas are not honored with O_DIRECT
680 Details : all writes with the flag O_DIRECT will use grants which leads to
681 this problem. Now using OBD_BRW_SYNC to guard this.
685 Bugzilla : 15713/16362
686 Description: Assertion in iopen_connect_dentry in 1.6.3
687 Details : looking up an inode via iopen with the wrong generation number can
688 populate the dcache with a disconneced dentry while the inode
689 number is in the process of being reallocated. This causes an
690 assertion failure in iopen since the inode's dentry list contains
691 both a connected and disconnected dentry.
695 Description: assertion failure in ldlm_handle2lock()
696 Details : fix a race between class_handle_unhash() and class_handle2object()
697 introduced in lustre 1.6.5 by bug 13622.
699 Severity : enhancement
701 Description: superblock lock contention with many SMP cores on one client
702 Details : several client filesystem locks were highly contended on SMP
703 NUMA systems with 8 or more cores. Per-CPU datastructures
704 and more efficient locking implemented to reduce contention.
709 Description: Kernel BUG: sd_iostats_bump: unexpected disk index
710 Details : remove the limit of 256 scsi disks in the sd_iostat patch
715 Description: oops in sd_iostats_seq_show()
716 Details : unloading/reloading the scsi low level driver triggers a kernel
717 bug when trying to access the sd iostat file.
722 Description: Kernel panics during QLogic driver reload
723 Details : REQ_BLOCK_PC requests are not handled properly in the sd iostat
724 patch, causing memory corruption.
729 Description: journal_dev option does not work in b1_6
730 Details : pass mount option during pre-mount.
732 Severity : enhancement
734 Description: Add a FIEMAP(FIle Extent MAP) ioctl for ldiskfs
735 Details : FIEMAP ioctl will allow an application to efficiently fetch the
736 extent information of a file. It can be used to map logical blocks
737 in a file to physical blocks in the block device.
740 Frequency : only with adaptive timeout enabled
742 Description: DEBUG_REQ() bad paging request
743 Details : ptlrpc_at_recv_early_reply() should not modify req->rq_repmsg
744 because it can be accessed by reply_in_callback() without the
748 Frequency : only on Cray X2
750 Description: X2 build failures
751 Details : fix build failures on Cray X2.
755 Description: xid & resent requests
756 Details : Initialize RPC XID from clock at startup (randomly if clock is
761 Description: quota recovery deadlock during mds failover
762 Details : This patch includes att18982, att18236, att18237 in bz14840.
764 1. fix osts hang when mds does failover with quotaon
765 2. prevent watchdog storm when osts threads wait for the
770 Description: kernel panic on racer
771 Details : Do not access dchild->d_inode when IS_ERR(dchild) is true.
773 Severity : enhancement
775 Description: Add lustre_start utility to start or stop multiple Lustre servers
780 Description: Lustre GPF in {:ptlrpc:ptlrpc_server_free_request+373}
781 Details : In case of memory pressure, list_del() can be called twice on
782 req->rq_history_list, causing a kernel oops.
786 Description: (ptllnd_peer.c:557:kptllnd_peer_check_sends()) ASSERTION(!in_interrupt()) failed
787 Details : fix stack overflow in the distributed lock manager by defering
788 export eviction after a failed ast to the elt thread instead of
789 handling it in the dlm interpret routine.
791 Severity : enhancement
793 Description: More exported tunables for mballoc
794 Details : Add support for tunable preallocation window and new tunables for
799 Description: Detect corruption of block bitmap and checking for preallocations
800 Details : Checks validity of on-disk block bitmap. Also it does better
801 checking of number of applied preallocations. When corruption is
802 found, it turns filesystem readonly to prevent further corruptions.
806 Frequency : only for big-endian servers
807 Description: Check for big-endian system while mounting fs with extents feature
808 Details : Mounting a filesystem with extents feature will fail on big-endian
809 systems since ext3-based ldiskfs is not supported on big-endian
810 systems. Can be over-ridden with "bigendian_extents" mount option.
814 Description: Excessive recovery window
815 Details : With AT enabled, the recovery window can be excessively long (6000+
816 seconds). To address this problem, we no longer use
817 OBD_RECOVERY_FACTOR when extending the recovery window (the connect
818 timeout no longer depends on the service time, it is set to
819 INITIAL_CONNECT_TIMEOUT now) and clients report the old service
820 time via pb_service_time.
824 Description: Watchdog triggered on MDS failover
825 Details : enable OBD_CONNECT_MDT flag when connecting from the MDS so that
826 the OSTs know that the MDS "UUID" can be reused for the same export
827 from a different NID, so we do not need to wait for the export to
830 -------------------------------------------------------------------------------
832 2008-05-26 Sun Microsystems, Inc.
834 * Support for kernels:
835 2.6.5-7.311 (SLES 9),
836 2.6.9-67.0.7.EL (RHEL 4),
837 2.6.16.54-0.2.5 (SLES 10),
838 2.6.18-53.1.14.el5 (RHEL 5),
839 2.6.22.14 vanilla (kernel.org)
840 * Client support for unpatched kernels:
841 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
842 2.6.16 - 2.6.22 vanilla (kernel.org)
843 * Due to problems with nested symlinks and FMODE_EXEC (bug 12652),
844 we do not recommend using patchless RHEL4 clients with kernels
845 prior to 2.6.9-55EL (RHEL4U5).
846 * Recommended e2fsprogs version: 1.40.7-sun1
847 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
848 * RHEL 4 and RHEL 5/SLES 10 clients behaves differently on 'cd' to a
849 removed cwd "./" (refer to Bugzilla 14399).
850 * A new quota file format has been introduced in 1.6.5.
851 The format conversion from prior releases is handled transparently,
852 but releases older than 1.4.12/1.6.5 will not understand this new
853 format. The automatic format conversion can be avoided by running
854 the following command on the MDS before upgrading:
855 'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
856 For more information, please refer to bugzilla 13904.
860 Description: quota performance fix
861 Details : quota data is written in journalled mode instead of ordered to
866 Description: lfs support for human-readable quota grace time strings
867 Details : lfs setquota -t and lfs quota -t represent quota grace times
868 in "XXwXXdXXhXXmXXs" format instead of large values in seconds
871 Frequency : always with o2ib 1.3 and sles10
873 Description: fix build with SLES10 and o2ib v3.
874 Details : sles10 uses diffrent name for Module.symver file but configure
875 assume this file has same name on RHEL/SLES/vanila kernels.
878 Frequency : very rare, if additional xattrs are used on kernels >= 2.6.12
880 Description: MDS may lose file striping (and hence file data) in some cases
881 Details : If there are additional extended attributes stored on the MDS,
882 in particular ACLs, SELinux, or user attributes (if user_xattr
883 is specified for the client mount options) then there is a risk
884 of attribute loss. Additionally, the Lustre file striping
885 needs to be larger than default (e.g. striped over all OSTs),
886 and an additional attribute must be stored initially in the
887 inode and then increase in size enough to be moved to the
888 external attribute block (e.g. ACL growing in size) for file
891 Severity : enhancement
893 Description: add message levels for liblustreapi
898 Description: MDT cannot be unmounted, reporting "Mount still busy"
899 Details : Mountpoint references were being leaked during open reply
900 reconstruction after an MDS restart. Drop mountpoint reference
901 in reconstruct_open() and free dentry reference also.
906 Description: fix for occasional failure case of -ENOSPC in recovery-small tests
907 Details : Move the 'good_osts' check before the 'total_bavail' check. This
908 will result in an -EAGAIN and in the exit call path we call
909 alloc_rr() which will with increasing aggressiveness attempt to
910 aquire precreated objects on the minimum number of required OSCs.
914 Description: Use old size assignment to avoid deadlock
915 Details : This reverts the changes in bugs 2369 and bug 14138 that introduced
916 the scheduling while holding a spinlock. We do not need locking
917 for size in ll_update_inode() because size is only updated from
918 the MDS for directories or files without objects, so there is no
919 other place to do the update, and concurrent access to such inodes
920 are protected by the inode lock.
924 Description: Use __u64 instead of int for valid bits
928 Description: resolve "_IOWR redefined" build error on SLES10
932 Description: dump the memory debugging after all modules are unloaded to
933 suppress false negative in conf_sanity test 39
937 Description: the recovery timer never expires
938 Details : for new client connect request, the recovery timer should not be
939 reset, otherwise recovery timer will never expired, if the old
940 client never come. Only old client connect and first connection
941 req should trigger recovery timer reset.
945 Description: the min numbers of lproc stats are wrong
946 Details : adding a new constant LC_MIN_INIT and use it for initialization
950 Frequency : always with interactive lfs
952 Description: Reinitialize optind to 0 so that interactive lfs works in all cases
955 Frequency : with multiple concurrent readdir processes in same directory
956 Bugzilla : 15406, 15169, 15175
957 Description: misc fixes for directory readahead.
958 Details : prevent previous statahead async RPC callback from processing the
959 current "statahead_info", race condition between async RPC callback
960 add dentry into dentry hash table and "ls" thread revalidate such
961 dentry, statahead his/miss control for hidden items, and so on.
963 Severity : enhancement
965 Description: build kernel-ib packages for OFED 1.3 in our release cycle
969 Description: incore types cleaning in quota code (with respect to 64-bit limits)
970 Details : several u32 variables declarations are replaced with u64 declarations
975 Description: fix SLES kernel versioning
976 Details : the kernel version for our SLES 10 kernel did not include a "-"
977 before the "smp" at the end. while this was not a problem in
978 general, it did mean that software trying to use the kernel
979 version to try to detect a vendor specific kernel would fail.
980 this was most evident by the OFED build scripts.
985 Description: Don't update lov_desc members until making sure they are valid
986 Details : When updating lov_desc members via proc fs, need fix their
987 validities before doing the real update.
990 Frequency : very rare
992 Description: don't put request into delay list while invalidate in flight.
993 Details : ptlrpc_delay_request sometimes put in delay list while invalidate
994 import in flight. this produce timeout for invalidate and sometimes
995 can cause stale data.
997 Severity : enhancement
999 Description: Update kernel to SLES9 2.6.5-7.311.
1001 Severity : enhancement
1003 Description: Update kernel to RHEL4 2.6.9-67.0.7.
1008 Frequency : on PPC only
1009 Description: not convert ost objects for directory because it's not exist.
1010 Details : ll_dir_getstripe assume dirrectory has ost objects but this wrong.
1012 Severity : enhancement
1014 Description: Fix warnings with compile liblustre at sles10/rhel5 which have
1015 __u64 as usingied long long type.
1018 Frequency : rare, on shutdown
1020 Description: race process ast vs remove callback
1021 Details : removing callback before disconnect import open race with
1022 processing callback.
1024 Severity : enhancement
1026 Description: Update kernel to SLES9 2.6.5-7.311.
1028 Severity : enhancement
1030 Description: Files open for execute are not marked busy on SLES10
1031 Details : Add FMODE_EXEC to SLES10 SP1 server kernel series.
1033 Severity : enhancement
1035 Description: Add server support for vanilla-2.6.22.14.
1038 Frequency : occasional
1040 Description: Avoid lov_create() getting stuck in obd_statfs_rqset()
1041 Details : If an OST is down the MDS will hang indefinitely in
1042 obd_statfs_rqset() waiting for the statfs data. While for
1043 MDS QOS usage of statfs, it should not stuck in waiting.
1045 Severity : enhancement
1047 Description: Disable adaptive timeouts by default
1050 Frequency : on network error
1052 Description: panic with double free request if network error
1053 Details : mdc_finish_enqueue is finish request if any network error occuring,
1054 but it's true only for synchronus enqueue, for async enqueue
1055 (via ptlrpcd) this incorrect and ptlrpcd want finish request
1059 Frequency : rare, on recovery
1061 Description: read procfs can produce deadlock in some situation
1062 Details : Holding lprocfs lock with send rpc can produce block for destroy
1063 obd objects and this also block reconnect with -EALREADY.
1064 This isn't fix all lprocfs bugs - but make it rare.
1066 Severity : enhancement
1068 Description: Update kernel to RHEL5 2.6.18-53.1.14.el5.
1071 Frequency : frequent on X2 node
1073 Description: mdc_set_open_replay_data LBUG
1074 Details : Set replay data for requests that are eligible for replay.
1079 Description: lustre_mgs: operation 101 on unconnected MGS
1080 Details : When MGC is disconnected from MGS long enough, MGS will evict the
1081 MGC, and late on MGC cannot successfully connect to MGS and a lot
1082 of the error messages complaining that MGS is not connected.
1085 Frequency : rare, depends on device drivers and load
1087 Description: MDS or OSS nodes crash due to stack overflow
1088 Details : Code changes in 1.6.4 increased the stack usage of some functions.
1089 In some cases, in conjunction with device drivers that use a lot
1090 of stack, the MDS (or possibly OSS) service threads could overflow
1091 the stack. One change which was identified to consume additional
1092 stack has been reworked to avoid the extra stack usage.
1094 Severity : enhancement
1096 Description: Update to RHEL5 latest kernel-2.6.18-53.1.13.el5.
1098 Severity : enhancement
1100 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.5.
1102 Severity : enhancement
1104 Description: Update to RHEL5 latest kernel-2.6.18-53.1.6.el5.
1106 Serverity : enhancement
1108 Description: Update RHEL4 kernel to 2.6.9-67.0.4.
1111 Frequency : rare on shutdown OST
1113 Description: Don't allow skipping OSTs if index has been specified.
1114 Details : Don't allow skipping OSTs if index has been specified, make locking
1115 in internal create lots better.
1120 Description: ASSERTION(!PageDirty(page)) failed
1121 Details : Wrong check could lead to an assertion failure under specific
1127 Description: LBUG in ptlrpc_check_set() bad phase ebc0de00
1128 Details : access to bitfield in structure is always rounded to long
1129 and this produce problem with not atomic change any bit.
1134 Description: Lustre make rpms failed.
1135 Details : Remove ldiskfs spec file to avoids rpmbuild be confused when
1136 builds Lustre rpms from tarball.
1138 Severity : enhancement
1140 Description: Update to SLES9 SP4 kernel-2.6.5-7.308.
1143 Frequency : rare on shutdown OST
1145 Description: If llog cancel was not send before clean_exports phase, this can
1146 produce deadlock in llog code.
1147 Details : If llog thread has last reference to obd and call class_import_put
1148 this produce deadlock because llog_cleanup_commit_master wait when
1149 last llog_commit_thread exited, but this never success because was
1150 called from llog_commit_thread.
1153 Frequency : only if OST index is skipped
1155 Description: NULL lov_tgts causing MDS oops
1156 Details : more safe checks for NULL lov_tgts for avoid oops.
1158 Severity : enhancement
1160 Description: Update to RHEL4 latest kernel-2.6.9-67.0.1.EL.
1162 Severity : enhancement
1164 Description: Update to RHEL5 latest kernel-2.6.18-53.1.4.el5.
1169 Description: make mgs_setparam() handle fsname containing dash
1170 Details : fsname containing a dash does not work with lctl conf_param
1172 Severity : enhancement
1174 Description: Update to RHEL4 Update-6 kernel-2.6.9-67.EL.
1177 Frequency : rare, in recovery and (or) destroy lovobjid file.
1179 Description: rewrite lov objid code.
1180 Details : Cleanup for lov objid code, remove scability problems and wrong
1181 locking. Fix sending last_id into ost.
1183 Severity : enhancement
1185 Description: Update to SLES10 SP1 latest kernel-2.6.16.54-0.2.3.
1187 Severity : enhancement
1189 Description: Update to RHEL5 Update-1 kernel 2.6.18-53.el5.
1190 Details : Use d_move_locked instead of __d_move.
1193 Frequency : rare, at shutdown
1195 Description: access already free / zero obd_namespace.
1196 Details : if client_disconnect_export was called without force flag set,
1197 and exist connect request in flight, this can produce access to
1198 NULL pointer (or already free pointer) when connect_interpret
1199 store ocd flags in obd_namespace.
1202 Frequency : only at startup
1204 Description: not alloc memory with spinlock held.
1205 Details : allocation memory with GFP_KERNEL can produce sleep deadlock,
1206 if any spinlock held.
1211 Description: lfs find does not continue on file error
1212 Details : Continue other files processing when a file/dir is absent.
1216 Description: Inconsistent usage of lustre_pack_reply()
1217 Details : Standardize the usage of lustre_pack_reply() such that it
1218 always generate a CERROR on failure.
1221 Frequency : very rare
1223 Description: Fix replay if there is an un-replied request and open
1224 Details : In some cases, older replay request will revert the
1225 mcd->mcd_last_xid on MDS which is used to record the client's
1226 latest sent request.
1228 Severity : enhancement
1230 Description: Update to RHEL5 kernel 2.6.18-8.1.15.el5.
1232 Severity : enhancement
1234 Description: Update to SLES10 SP1 kernel 2.6.16.53-0.16
1236 Severity : enhancement
1238 Description: Update to SLES9 kernel-2.6.5-7.287.3.
1240 Severity : enhancement
1242 Description: Update to RHEL4 kernel-2.6.9-55.0.12.EL.
1244 Severity : enhancement
1246 Description: Build SLES10 patchless client fails
1247 Details : The configure was broken by run ./configure with
1248 --with-linux-obj=.... argument for patchless client. When the
1249 configure use --with-linux-obj, the LINUXINCLUDE= -Iinclude
1250 can't search header adequately. Use absolute path such as
1251 -I($LINUX)/include instead.
1253 Severity : enhancement
1255 Description: Lustre Page Accounting
1256 Details : New macros for page alloc and free which enable accounting
1257 of page allocation of Lustre. Use percpu counters to store memory
1258 and page statistics.
1261 Frequency : only if debugging is disabled
1263 Description: LASSERT_{REQ,REP}SWAB macros are buggy
1264 Details : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
1265 LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
1266 these macros and replace them with their definitions instead.
1271 Description: interrupt oig_wait produce painc on resend.
1272 Details : brw_redo_request can be used for resend requests from ptlrpcd and
1273 private set, and this produce situation when rq_ptlrpcd_data not
1274 copyed to new allocated request and triggered LBUG on assert
1275 req->rq_ptlrpcd_data != NULL. But this member used only for wakeup
1276 ptlrpcd set if request is changed and can be safety changed to use
1279 Severity : enhancement
1281 Description: organize the server-side client stats on per-nid basis
1282 Details : Change the structure of stats under obdfilter and mds to
1290 The "uuid"s file would list the uuids of _active_ exports.
1291 And the clear entry is to clear all stats and stale nids.
1296 Description: Processes looping in ll_readdir() on Lustre clients finally causing
1297 a full node pseudo-hang
1298 Details : Concurrent access to the same directory from multiple clients with
1299 intensive file creation/removal can cause a client node to spin in
1300 ll_readdir(). i_version must be increased every time the lock is
1301 cancelled to ensure a revalidate is done.
1306 Description: touch file failed when fs is not full
1307 Details : OST in recovery should not be discarded by MDS in alloc_qos(),
1308 otherwise we can get ENOSP while fs is not full.
1311 Frequency : only for Cray XT3
1312 Bugzilla : 12829/13455
1313 Description: Changing primary group doesn't change the group lustre assigns to
1315 Details : When CRAY_XT3 is defined, the fsgid supplied by the client is
1316 overridden with the primary group provided by the group upcall,
1317 whereas the supplied fsgid can be trusted if it is in the list of
1318 supplementary groups returned by the group upcall.
1320 Severity : enhancement
1322 Description: Root Squash Functionality
1323 Details : Implementation of NFS-like root squash capability. Specifically,
1324 don't allow someone with root access on a client node to be able
1325 to manipulate files owned by root on a server node.
1327 Severity : enhancement
1329 Description: Slow trucate/writes to huge files at high offsets.
1330 Details : Directly associate cached pages to lock that protect those pages,
1331 this allows us to quickly find what pages to write and remove
1332 once lock callback is received.
1337 Description: Too many locks accumulating on client during NFS usage
1338 Details : mds_open improperly used accmode to find out access mode to a
1339 file. Also mdc_intent_lock logic to find out if we already have
1340 lock similar to just received was flawed since introduction of
1341 skiplists - locks are now added to the front of the granted
1346 Description: Hit ASSERTION(obd->obd_stopping == 1) failed in some setup failed
1348 Details : In obd setup failure handler, obd_stopping will not necessarily to
1349 be 1, and obd_set_up should also be checked to make sure whether
1350 obd is completely setup.
1352 Severity : enhancement
1354 Description: Allow masking D_WARNING, D_ERROR messages from console
1355 Details : Console messages can now be disabled via lnet.printk.
1360 Description: User code with malformed file open parameter crashes client node
1361 Details : Before packing join_file req, all the related reference should be
1362 checked carefully in case some malformed flags cause fake join_file
1368 Description: shrink/enlarge qunit size when needed; fix the problem of coarse
1369 grain of quota doing harm to quota's accuracy
1370 Details : qunit size will be changed when quota limitation is too low/high;
1371 record the pending quota write in order to get more accureate
1372 quota; delete the patch for bug12588, which is unnecessary when
1373 this patch is landed. This bug also contains fixes for bug 14526,
1374 14299, 14601 and 13794.
1378 Description: LDLM_ENQUEUE races with LDLM_CP_CALLBACK
1379 Details : ldlm_completion_ast() assumes that a lock is granted when the req
1380 mode is equal to the granted mode. However, it should also check
1381 that LDLM_FL_CP_REQD is not set.
1385 Description: Heavy nfs access might result in deadlocks
1386 Details : After ELC code landed, it is now improper to enqueue any mds
1387 locks under och_sem, because enqueue might want to decide to
1388 cancel open locks for same inode we are holding och_sem for.
1392 Description: 35% write performance drop with ldiskfs2 when quotas are on
1393 Details : Enable ext3 journalled quota by default to improve performance
1394 when quotas are turned on.
1398 Description: Client eviction while running blogbench
1399 Details : A lot of unlink operations with concurrent I/O can lead to a
1400 deadlock causing evictions. To address the problem, the number of
1401 oustanding OST_DESTROY requests is now throttled to
1402 max_rpcs_in_flight per OSC and LDLM_FL_DISCARD_DATA blocking
1403 callbacks are processed in priority.
1406 Frequency : RHEL4 only
1408 Description: mkfs is very slow on IA64/RHEL4
1409 Details : A performance regression has been discovered in the MPT Fusion
1410 driver between versions 3.02.73rh and 3.02.99.00rh. As a
1411 consequence, we have downgraded the MPT Fusion driver in the RHEL4
1412 kernel from 3.02.99.00 to 3.02.73 until this problem is fixed.
1415 Frequency : PPC/PPC64 only
1417 Description: conflicts between asm-ppc64/types.h and lustre_types.h
1418 Details : fix duplicated definitions between asm-ppc64/types.h and
1419 lustre_types.h on PPC.
1422 Frequency : PPC/PPC64 only
1424 Description: asm-ppc/segment.h does not exist
1425 Details : fix compile issue on PPC.
1429 Description: data checksumming impacts single node performance
1430 Details : add support for several checksum algorithms. Currently, CRC32 and
1431 Adler-32 are supported. The checksum type can be changed on the fly
1432 through /proc/fs/lustre/osc/*/checksum_type.
1436 Description: use adler32 for page checksums
1437 Details : when available, use the Adler-32 algorithm instead of CRC32 for
1442 Description: better handle error messages in extents code
1444 Severity : enhancement
1446 Description: SNMP support enhancement
1447 Details : Adding total number of sampled request for an MDS node in snmp
1450 Severity : enhancement
1452 Description: Optimize ldlm waiting list processing for PR extent locks
1453 Details : When processing waiting list for read extent lock and meeting read
1454 lock that is same or wider to it that is not contended, skip
1455 processing rest of the list and immediatelly return current
1456 status of conflictness, since we are guaranteed there are no
1457 conflicting locks in the rest of the list.
1461 Description: Time out and refuse to reconnect
1462 Details : When the failover node is the primary node, it is possible
1463 to have two identical connections in imp_conn_list. We must
1464 compare not conn's pointers but NIDs, otherwise we
1465 can defeat connection throttling.
1469 Description: Client not clear own cache if answer to reconnect is lost.
1470 Details : Client gets evicted from server. Now client also thinks it is
1471 disconnected (or gets ENOTCONN on its operation) and decides to
1472 reconnect. Server receives reconnect message, but cannot find
1473 export. New export is created that is fully valid (new cookie!),
1474 but reply is lost and not reported to client. Client reconnects
1475 again and gets back a just-created connection, but it is not new
1476 so client thinks it was not evicted and does not do recovery.
1480 Description: Detect stride IO mode in read-ahead
1481 Details : When a client does stride read, read-ahead should detect that and
1482 read-ahead pages according to the detected stride pattern.
1486 Description: build for x2 fails
1487 Details : fix compile issue on Cray systems.
1489 Severity : enhancement
1491 Description: implement readv/writev APIs(aio_read/aio_writes in newer kernels)
1492 Details : This greatly improves speed of NFS writes on 2.6 kernels.
1495 Frequency : only on PPC/SLES10
1497 Description: "BITS_PER_LONG is not 32 or 64" in linux/idr.h
1498 Details : On SLES10/PPC, fs.h includes idr.h which requires BITS_PER_LONG to
1499 be defined. Add a hack in mkfs_lustre.c to work around this compile
1504 Description: LASSERT on MDS when client holding flock lock dies
1505 Details : ldlm pool logic depends on number of granted locks equal to
1506 number of released locks which is not true for flock locks, so
1507 just exclude such locks from consideration.
1511 Description: MDS deadlock with many ll_sync_lov threads and I/O stalled
1512 Details : Use fsfilt_sync() for both the whole filesystem sync and
1513 individual file sync to eliminate dangerous inode locking
1514 with I_LOCK that can lead to a deadlock.
1518 Description: Update an obsolete wirecheck.c generator
1519 Details : Update wirecheck.c/wirehdr.c and regenerate wiretest.c
1523 Description: Client can panic on open sometimes
1524 Details : It is possible that we try to free already freed request in
1525 ll_file_open in some error cases when we send request from
1530 Description: performance in 1.6.3
1531 Details : Force q->max_phys_segments to MAX_PHYS_SEGMENTS on SLES10 to be
1532 sure that 1MB requests are not fragmented by the block layer.
1536 Description: LDLM soft lockups - improvement
1537 Details : It is be possible to send the lock handle along with each read
1538 or write request because the client is already doing a lock match
1539 itself so there isn't any reason the OST should have to re-do that
1545 Description: lfs quota fails with deactivated OSTS
1546 Details : With this patch, three improvements are included:
1547 1. detete the softlimit in mds and osts when use "lfs quota".
1548 2. display the inaccurate data in the output of "lfs quota".
1549 3. try to get quota info when "lfs quota" is executed.
1554 Description: Extent locks not granted with no conflicts sometimes.
1555 Details : When race occurs in glimpse handler and nothing is returned,
1556 we do not reprocess the queue after lock cancel, and that leads
1557 to a stall until next activity on a resource
1560 Frequency : failover with quotaon
1562 Description: during mds failovers with quota on, OSTs got into deadlock state
1563 and causing dumpstack.
1564 Details : for every quota slave, at any time, there is only one quota req
1565 is sent to quota master for every uid/gid. Before that quota req
1566 returns, all the thread relative to the same uid/gid will wait.
1567 So if the quota req is lost because mds failovers or any other
1568 reasons, this bug will be hit. Now, dqacq_interpret() will handle
1569 quota reqs who time out.
1571 Severity : enhancement
1574 Description: when quota slave checks if quota is enough, there is an unnecessary
1576 Details : place this wait on necessary place instead of always waiting.
1578 --------------------------------------------------------------------------------
1580 2007-12-07 Cluster File Systems, Inc. <info@clusterfs.com>
1582 * Support for kernels:
1583 2.6.5-7.286 (SLES 9),
1584 2.6.9-55.0.9.EL (RHEL 4),
1585 2.6.16.53-0.8 (SLES 10),
1586 2.6.18-8.1.14.el5 (RHEL 5),
1587 2.6.18.8 vanilla (kernel.org)
1588 * Client support for unpatched kernels:
1589 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1590 2.6.16 - 2.6.22 vanilla (kernel.org)
1591 * Due to recently discovered recovery problems, we do not recommend
1592 using patchless RHEL 4 clients with this or any earlier release.
1593 * Recommended e2fsprogs version: 1.40.2-cfs1
1594 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1597 Frequency : occasional
1599 Description: MDS hang or stay in waiting lock
1600 Details : If client receive lock with CBPENDING flag ldlm need send lock
1601 cancel as separate rpc, to avoid situation when cancel request
1602 can't processed due all i/o threads stay in wait lock.
1605 Frequency : occasional
1607 Description: Do not fail import if osc_interpret_create gets -EAGAIN
1608 Details : If osc_interpret_create got -EAGAIN it immediately exits and
1609 wakeup oscc_waitq. After wakeup oscc_wait_for_objects call
1610 oscc_has_objects and see OSC has no objests and call
1611 oscc_internal_create to resend create request.
1613 Severity : enhancement
1615 Description: Update kernel patches for SLES10 2.6.16.53-0.8.
1616 Details : Update which_patch & target file for SLES10 latest kernel.
1618 Severity : enhancement
1620 Description: add --type and --size parameters to lfs find
1621 Details : Enhance lfs find by adding filetype and filesize parameters. Also
1622 multiple OBDs can now be specified for the --obd option.
1624 Severity : enhancement
1626 Description: eliminate client locks in face of contention
1627 Details : file contention detection and lockless i/o implementation
1628 for contended files.
1630 Severity : enhancement
1632 Description: Remove client patches from SLES 10 kernel.
1633 Details : This causes SLES 10 clients to behave as patchless clients
1634 even on a Lustre-patched (server) kernel.
1636 Severity : enhancement
1638 Description: use i_size_read and i_size_write in 2.6 port
1639 Details : replace inode->i_size access with i_size_read/write()
1642 Frequency : when removing large files
1644 Description: scheduling issue during removal of large Lustre files
1645 Details : Don't take the BKL in fsfilt_ext3_setattr() for 2.6 kernels.
1646 It causes scheduling issues when removing large files (17TB in the
1652 Description: 1.4.11 Can't handle directories with stripe set and extended ACLs
1653 Details : Impossible (EPROTO is returned) to access a directory that has a
1654 non-default striping and ACLs.
1657 Frequency : only on ppc
1659 Description: /proc/fs/lustre/devices broken on ppc
1660 Details : The patch as applied to 1.6.2 doesn't look correct for all arches.
1661 We should make sure the type of 'index' is loff_t and then cast
1662 explicitly as needed below. Do not assign an explicitly cast
1666 Frequency : only for rhel5
1668 Description: Kernel patches update for RHEL5 2.6.18-8.1.10.el5.
1669 Details : Modify the target file & which_kernel.
1672 Frequency : if the uninit_groups feature is enabled on ldiskfs
1674 Description: e2fsck reports "invalid unused inodes count"
1675 Details : If a new ldiskfs filesystem is created with the "uninit_groups"
1676 feature and only a single inode is created in a group then the
1677 "bg_unused_inodes" count is incorrectly updated. Creating a
1678 second inode in that group would update it correctly.
1683 Description: buffer overruns could theoretically occur
1684 Details : llapi_semantic_traverse() modifies the "path" argument by
1685 appending values to the end of the origin string, and a buffer
1686 overrun may occur. Adding buffer overrun check in liblustreapi.
1688 Severity : enhancement
1690 Description: Add jbd statistics patch for RHEL5 and 2.6.18-vanilla.
1693 Frequency : only if filesystem is inconsistent
1695 Description: handle "serious error: objid * already exists" more gracefully
1696 Details : If LAST_ID value on disk is smaller than the objects existing in
1697 the O/0/d* directories, it indicates disk corruption and causes an
1698 LBUG(). If the object is 0-length, then we should use the existing
1699 object. This will help to avoid a full fsck in most cases.
1701 Severity : enhancement
1703 Description: Kernel patches update for RHEL4 2.6.9-55.0.6.
1704 Details : Modify vm-tunables-rhel4.patch.
1706 Severity : enhancement
1708 Description: Kernel config for 2.6.18-vanilla.
1709 Details : Modify targets/2.6-vanilla.target.in.
1710 Add config file kernel-2.6.18-2.6-vanilla-i686.config.
1711 Add config file kernel-2.6.18-2.6-vanilla-i686-smp.config.
1712 Add config file kernel-2.6.18-2.6-vanilla-x86_64.config.
1713 Add config file kernel-2.6.18-2.6-vanilla-x86_64-smp.config.
1716 Frequency : occasional
1718 Description: improve handling recoverable errors
1719 Details : If request processed with error which can be recoverable on server
1720 request should be resend, otherwise page released from cache and
1726 Description: Kernel patches update for RHEL5 2.6.18-8.1.14.el5.
1727 Details : Modify target file & which_patch.
1728 A flaw was found in the IA32 system call emulation provided
1729 on AMD64 and Intel 64 platforms. An improperly validated 64-bit
1730 value could be stored in the %RAX register, which could trigger an
1731 out-of-bounds system call table access. An untrusted local user
1732 could exploit this flaw to run code in the kernel
1733 (ie a root privilege escalation). (CVE-2007-4573).
1737 Description: change order of libsysio includes
1738 Details : '#include sysio.h' should always come before '#include xtio.h'
1740 Severity : enhancement
1742 Description: adapt the lustre_config script to support the upgrade case
1743 Details : Add "-u" option for lustre_config script to support upgrading 1.4
1744 server targets to 1.6 in parallel.
1749 Description: To avoid grant space > avaible space when the disk is almost
1750 full. Without this patch you might see the error "grant XXXX >
1751 available" or some LBUG about grant, when the disk is almost
1753 Details : In filter_check_grant, for non_grant cache write, we should
1754 check the left space by if (*left > ungranted + bytes), instead
1755 of (*left > ungranted), because only we are sure the left space
1756 is enough for another "bytes", then the ungrant space should be
1757 increase. In client, we should update cl_avail_grant only there is
1758 OBD_MD_FLGRANT in the reply.
1763 Description: Update RHEL 4 kernel to fix local root privilege escalation.
1764 Details : Update to the latest RHEL 4 kernel to fix the vulnerability
1765 described in CVE-2007-4573. This problem could allow untrusted
1766 local users to gain root access.
1769 Frequency : when using O_DIRECT and quotas
1771 Description: Incorrect file ownership on O_DIRECT output files
1772 Details : block usage reported by 'lfs quota' does not take into account
1773 files that have been written with O_DIRECT.
1777 Description: (rw.c:1323:ll_read_ahead_pages()) ASSERTION(page_idx > ria->ria_stoff) failed
1778 Details : Once the unmatched stride IO mode is detected, shrink the stride-ahead
1779 window to 0. If it does hit cache miss, and read-pattern is still
1780 stride-io mode, does not reset the stride window, but also does not
1781 increase the stride window length in this case.
1783 --------------------------------------------------------------------------------
1785 2007-09-27 Cluster File Systems, Inc. <info@clusterfs.com>
1787 * Support for kernels:
1788 2.6.5-7.286 (SLES 9),
1789 2.6.9-55.0.2.EL (RHEL 4),
1790 2.6.16.46-0.14 (SLES 10),
1791 2.6.18-8.1.8.el5 (RHEL 5),
1792 2.6.18.8 vanilla (kernel.org)
1793 * Client support for unpatched kernels:
1794 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
1795 2.6.16 - 2.6.21 vanilla (kernel.org)
1796 * Due to recently discovered recovery problems, we do not recommend
1797 using patchless RHEL 4 clients with this or any earlier release.
1798 * Recommended e2fsprogs version: 1.40.2-cfs1
1799 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
1803 Description: Fix errors in lfs documentation
1804 Details : Fixes man pages
1806 Severity : enhancement
1808 Description: Adaptive timeouts
1809 Details : RPC timeouts adapt to changing server load and network
1810 conditions to reduce resend attempts and improve recovery time.
1812 Severity : enhancement
1814 Description: llapi_file_create() does not allow some changes
1815 Details : add llapi_file_open() that allows specifying the file creation
1816 mode and open flags, and also returns an open file handle.
1818 Severity : enhancement
1820 Description: df doesn't work properly if diskfs blocksize != 4K
1821 Details : Choose biggest blocksize of OST's as the LOV's blocksize.
1824 Frequency : always on directories with default striping set
1826 Description: lfs find on -1 stripe looping in lsm_lmm_verify_common()
1827 Details : Avoid lov_verify_lmm_common() on directory with -1 stripe count.
1829 Severity : enhancement
1831 Description: merge and cleanup kernel patches.
1832 Details : Remove mnt_lustre_list in vfs_intent-2.6-rhel4.patch.
1834 Severity : enhancement
1836 Description: RedHat Update kernel for RHEL5
1837 Details : Kernel config file for RHEL5.
1839 Severity : enhancement
1841 Description: OSS needs mutliple precreate threads
1842 Details : Add ability to start more than one create thread per OSS.
1845 Frequency : only with quota on the root user
1847 Description: mds_obd_create error creating tmp object
1848 Details : When the user sets quota on root, llog will be affected and can't
1849 create files and write files.
1852 Frequency : Always on ia64 patchless client, and possibly others.
1854 Description: Add EXPORT_SYMBOL check for node_to_cpumask symbol.
1855 Details : This allows the patchless client to be loaded on architectures
1856 without this export.
1861 Description: disorder of journal start and llog_add cause deadlock.
1862 Details : in llog_origin_connect, journal start should happen before
1863 llog_add keep the same order as other functions to avoid
1866 Severity : enhancement
1868 Description: RedHat Update kernel for RHEL5
1869 Details : Modify the kernel config file more closer RHEL5.
1871 Severity : enhancement
1873 Description: Build failure against Centos5 (RHEL5)
1874 Details : Define PAGE_SIZE when it isn't present.
1877 Frequency : occasionally when using NFS
1879 Description: "ll_intent_file_open()) lock enqueue: err: -13" with nfs
1880 Details : with NFS, the anon dentry's parent was set to itself in
1881 d_alloc_anon(), so in MDS, we use rec->ur_fid1 to find the
1882 corresponding dentry other than use rec->ur_name.
1884 Severity : enhancement
1886 Description: enable data checksumming by default
1887 Details : enable checksum by default, allow --disable-checksum
1888 configure option and "-o nochecksum" mount option. Checksums
1889 can also be disabled at runtime via $LPROC/osc/*/checksum_pages.
1892 Frequency : Occasionally with failover
1894 Description: Client eviction due to failover config
1895 Details : after a connection loss, the lustre client should attempt to
1896 reconnect to the last active server first before trying the
1897 other potential connections.
1899 Severity : enhancement
1901 Description: client-side metadata stat-ahead during readdir(directory readahead)
1902 Details : perform client-side metadata stat-ahead when the client detects
1903 readdir and sequential stat of dir entries therein
1908 Description: when mds and osts use different quota unit(32bit and 64bit),
1909 quota will be released repeatly.
1910 Details : void sending multiple quota reqs to mds, which will keep the
1911 status between the reqs.
1914 Frequency : only with liblustre clients on XT3
1916 Description: evictions taking too long
1917 Details : allow llrd to evict clients directly on OSTs
1921 Description: osts not allocated evenly to files
1922 Details : change the condition to increase offset_idx
1924 Severity : enhancement
1926 Description: self-adjustable client's lru lists
1927 Details : use adaptive algorithm for managing client cached locks lru
1928 lists according to current server load, other client's work
1929 pattern, memory activities, etc. Both, server and client
1930 side namespaces provide number of proc tunables for controlling
1935 Description: rewrite ext2-derived code in llite/dir.c and obdclass/uuid.c
1936 Details : rewrite inherited code (uuid parsing code from ext2 utils and
1937 readdir code from ext3) from scratch preserving functionality.
1941 Description: Only those disconnect error should be returned by rq_status.
1942 Details : In open/enqueue processs, Some errors, which will cause client
1943 disconnected, should be returned by rq_status, while other
1944 errors should still be returned by intent, then mdc or llite will
1947 Severity : enhancement
1949 Description: Tune the kernel for good SCSI performance.
1950 Details : Set the value of /sys/block/{dev}/queue/max_sectors_kb
1951 to the value of /sys/block/{dev}/queue/max_hw_sectors_kb
1955 Frequency : Always for filesystems larger than 2TB on 32-bit systems.
1956 Bugzilla : 13547, 13627
1957 Description: Data corruption for OSTs that are formatted larger than 2TB
1959 Details : When generating the bio request for lustre file writes the
1960 sector number would overflow a temporary variable before being
1961 used for the IO. The data reads correctly from Lustre (which
1962 will overflow in a similar manner) but other file data or
1963 filesystem metadata may be corrupted in some cases.
1967 Description: "lfs find -obd UUID" prints directories
1968 Details : "lfs find -obd UUID" will return all directory names instead
1969 of just file names. It is incorrect because the directories
1970 do not reside on the OSTs.
1973 Bugzilla : 13596, 13614
1974 Description: MDS hang after unclean shutdown of lots of clients
1975 Details : Never resend AST requests.
1979 Frequency : Always, for kernels after 2.6.16
1980 Description: Fix warning idr_remove called for id=.. which is not allocated.
1981 Details : Last kernels save old s_dev before kill super and not allow
1982 to restore from callback - restore it before call kill_anon_super.
1986 Description: Fix error on 'ls .' at the top of the Lustre mount.
1987 Details : Don't revalidate dentry if it is a root dentry.
1989 --------------------------------------------------------------------------------
1991 2007-08-27 Cluster File Systems, Inc. <info@clusterfs.com>
1993 * Support for kernels:
1994 2.6.5-7.286 (SLES 9),
1995 2.6.9-55.0.2.EL (RHEL 4),
1996 2.6.16.46-0.14 (SLES 10),
1997 2.6.18-8.1.8.el5 (RHEL 5),
1998 2.6.18.8 vanilla (kernel.org)
1999 * Client support for unpatched kernels:
2000 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2001 2.6.16 - 2.6.22 vanilla (kernel.org)
2002 * Due to recently discovered recovery problems, we do not recommend
2003 using patchless RHEL 4 clients with this or any earlier release.
2004 * Recommended e2fsprogs version: 1.39.cfs8
2005 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2010 Description: block reactivating mgc import until all deactivates complete
2011 Details : Fix race when failing back MDT/MGS to itself (testing)
2013 Severity : enhancement
2015 Description: lfs setstripe enhancement
2016 Details : Make lfs setstripe understand 'k', 'm' and 'g' for stripe size.
2019 Frequency : mds/oss recovery
2021 Description: llog ctxt is refrenced after it has been freed.
2022 Details : llog ctxt refcount was added to avoide the race between ctxt free
2023 and llog recovery process. Each llog user must hold ctxt refcount
2024 before it access the llog. And the llog ctxt can only be freed
2025 when its refcount is zero.
2027 Severity : enhancement
2029 Description: randomly memory allocation failure util
2030 Details : Make lustre randomly failed allocating memory for testing purpose.
2033 Frequency : only for SLES10
2035 Description: Update kernel patch for SLES10 SP1
2036 Details : Add patch blkdev_tunables-2.6-sles10.patch to 2.6-sles10.series.
2038 Severity : enhancement
2040 Description: omit set fsid for export NFS
2041 Details : fix set/restore device id for avoid EMFILE error and mark lustre fs
2042 as FS_REQUIRES_DEV for avoid problems with generate fsid.
2045 Frequency : after network failures
2047 Description: Add sync option to mount_lustre.c
2048 Details : Client loses data written to lustre after a network interruption.
2050 Severity : enhancement
2052 Description: Error message improvement.
2053 Details : Merging of two LCONSOLE_ERROR_MSG into one.
2055 Severity : enhancement
2057 Description: don't use GFP_* in generic Lustre code.
2058 Details : Use cfs_alloc_* functions and CFS_* flags for code portability.
2060 Severity : enhancement
2062 Description: obdclass is limited by single OBD_ALLOC(idarray)
2063 Details : replace OBD_ALLOC/OBD_FREE with OBD_VMALLOC/OBD_VFREE
2065 Severity : enhancement
2067 Description: updated patchess for new RHEL4 kernel
2068 Details : Fixed ext3-unlink-race.patch per Kalpak's comment.
2070 Severity : enhancement
2072 Description: warnings with build patchless client with vanila 2.6.19 and up
2073 Details : change old ctl_table style and replace ctl_table/ctl_table_header
2074 with cfs_sysctl_table_t/cfs_sysctl_table_header_t
2076 Severity : enhancement
2078 Description: O_DIRECT bypasses client statistics.
2079 Details : When running with O_DIRECT I/O, neither the client rpc_stats nor
2080 read_ahead_stats were updated. Copied stats section from
2081 osc_send_oap_rpc() into async_internal().
2084 Frequency : only for Cray XT3
2086 Description: peer credits not enough on many OST per OSS systems.
2087 Details : Use new lnet way to add credits as we need those for pings and ASTs
2090 Frequency : only with liblustre
2092 Description: Liblustre is not releasing flock locks on file close.
2093 Details : Release flock locks on file close.
2096 Frequency : only for RHEL4
2098 Description: Update kernel patches for kernel-2.6.9-55.0.2.EL
2099 Details : Remove inode-nr_unused-2.6.9-rhel4.patch from 2.6-rhel4.series
2100 Update target file and kernel config.
2104 Description: lustre support for RHEL5
2105 Details : Add support for RHEL5.
2110 Description: ASSERTION(export != NULL) failed in target_handle_connect
2111 Details : Assetion hit is result of rare race between disconnect and connet
2112 to same nid. target_handle_connect found old connect cockie and
2113 tried to reconnect, but can't find export for this cockie.
2118 Description: umount blocks forever on error
2119 Details : In result of wrong using obd_no_recov and obd_force flags client
2120 can hand if cancel or some other requests is lost.
2123 Frequency : Only for SLES
2125 Description: sanity_quota fail test_1
2126 Details : There are multiple occurences of $TSTUSR in SLES's /etc/group
2127 file, which makes TSTID[2] inunique.
2129 Severity : enhancement
2131 Description: Kernel patches for SLES9 2.6.5-7.286 kernel
2132 Details : Update target/ChangeLog/which_patch .
2134 Severity : enhancement
2136 Description: jbd statistics
2137 Details : Port older jbd statistics patch for sles10
2142 Description: lvbo_init failed for resource with missing objects.
2143 Details : Fix returning error if we do stat for file with missing/corrupted
2144 objects and i_size set to all sum of size of all avaible objects.
2145 if we to truncate/write to missing object - it is recreated.
2150 Description: Oops in read and write path when failing to allocate lock.
2151 Details : Check if lock allocation failed and return error back.
2154 Frequency : When flocks are used.
2156 Description: assertion failure in ldlm_cli_enquque_fini for non NULL lock.
2157 Details : Flock locks might destroy just granted lock if it could be merged
2158 with another existing flock, this is done in completion handler,
2159 so teach ldlm_cli_enquque_fini that this is a valid case for
2165 Description: reply_lock_interpret crash due to race with it and lock cancel.
2166 Details : Do not replay locks that are being cancelled. Do not reference
2167 locks by their address during replay, just by their handle.
2170 Frequency : only with deactivated OSTs
2172 Description: lstripe command fails for valid OST index
2173 Details : The stripe offset is compared to 'lov->desc.ld_tgt_count'
2174 instead of lov->desc.ld_active_tgt_count.
2176 Severity : enhancement
2178 Description: Build failure against Centos5 (RHEL5)
2179 Details : Use getpagesize() instead of PAGE_SIZE.
2181 --------------------------------------------------------------------------------
2183 2007-07-30 Cluster File Systems, Inc. <info@clusterfs.com>
2185 * Support for kernels:
2186 2.6.5-7.283 (SLES 9),
2187 2.6.9-55.EL (RHEL 4),
2188 2.6.16.46-0.14 (SLES 10),
2189 2.6.18.8 vanilla (kernel.org)
2190 * Client support for unpatched kernels:
2191 (see http://wiki.lustre.org/index.php?title=Patchless_Client)
2192 2.6.16 - 2.6.22 vanilla (kernel.org)
2193 * Due to recently discovered recovery problems, we do not recommend
2194 using patchless RHEL 4 clients with this or any earlier release.
2195 * Recommended e2fsprogs version: 1.39.cfs8
2196 * Note that reiserfs quotas are disabled on SLES 10 in this kernel.
2197 * Starting with this release, the ldiskfs backing filesystem required
2198 by Lustre is now in its own package, lustre-ldiskfs. This package
2199 should be installed. It is versioned separately from Lustre and
2200 may be released separately in future.
2205 Description: server LBUG when shutting down
2206 Details : Block umount forever until the mount refcount is zero rather
2207 than giving up after an arbitrary timeout.
2209 Severity : enhancement
2211 Description: add optional extra BUILD_VERSION info
2212 Details : add a new environment variable (namely LUSTRE_VERS) which allows
2213 to override the lustre version.
2216 Frequency : 2.6.18 servers only
2218 Description: ll_kern_mount() doesn't release the module reference
2219 Details : The ldiskfs module reference count never drops down to 0
2220 because ll_kern_mount() doesn't release the module reference.
2225 Description: server LBUG when using old ost_num_threads parameter
2226 Details : Accept the old ost_num_threads parameter but warn that it
2227 is deprecated, and fix an off-by-one error that caused an LBUG.
2232 Description: Transient SCSI error results in persistent IO issue
2233 Details : iobuf->dr_error is not reinitialized to 0 between two
2237 Frequency : sometimes when underlying device returns I/O errors
2239 Description: OSTs not going read-only during write failures
2240 Details : OSTs are not remounted read-only when the journal commit threads
2241 get I/O errors because fsfilt_ext3 calls journal_start/stop()
2242 instead of the ext3 wrappers.
2245 Frequency : only with 10000 clients or more
2247 Description: poor connect scaling with increasing client count
2248 Details : Don't run filter_grant_sanity_check for more than 100 exports
2249 to improve scaling for large numbers of clients.
2252 Frequency : SLES10 only
2254 Description: sanity-quota.sh quotacheck failed: rc = -22
2255 Details : Quotas cannot be enabled on SLES10.
2258 Frequency : liblustre clients only
2260 Description: getdirentries does not give error when run on compute nodes
2261 Details : getdirentries does not fail when the size specified as an argument
2262 is too small to contain at least one entry
2264 Severity : enhancement
2266 Description: Add LNET router traceability for debug purposes
2267 Details : If a checksum failure occurs with a router as part of the
2268 IO path, the NID of the last router that forwarded the bulk data
2269 is printed so it can be identified.
2274 Description: OST "spontaneously" evicts client; client has imp_pingable == 0
2275 Details : Due to a race condition, liblustre clients were occasionally
2276 evicted incorrectly.
2278 Severity : enhancement
2280 Description: lfs setstripe use optional parameters instead of postional
2283 Severity : enhancement
2285 Description: Nanosecond timestamp support for ldiskfs
2286 Details : The on-disk ldiskfs filesystem has added support for nanosecond
2287 resolution timestamps. There is not yet support for this at
2288 the Lustre filesystem level.
2291 Frequency : during server recovery
2293 Description: MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING
2294 Details : request with rq_no_resend flag not awake l_wait_event if they get a
2298 Frequency : nfs export on patchless client
2300 Description: connectathon hang when test nfs export over patchless client
2301 Details : Disconnected dentry cannot be found with lookup, so we do not need
2302 to unhash it or make it invalid
2305 Description: fix llapi_lov_get_uuids() to allow many OSTs to be returned
2306 Details: : Change llapi_lov_get_uuids() to read the UUIDs from /proc instead
2307 of using an ioctl. This allows lfsck for > 160 OSTs to succeed.
2312 Description: open req refcounting wrong on reconnect
2313 Details : If reconnect happened between getting open reply from server and
2314 call to mdc_set_replay_data in ll_file_open, we will schedule
2315 replay for unreferenced request that we are about to free.
2316 Subsequent close will crash in variety of ways.
2317 Check that request is still eligible for replay in
2318 mdc_set_replay_data().
2323 Description: disable writes to filesystem when reading health_check file
2324 Details : the default for reading the health_check proc file has changed
2325 to NOT do a journal transaction and write to disk, because this
2326 can cause reads of the /proc file to hang and block HA state
2327 checking on a healthy but otherwise heavily loaded system. It
2328 is possible to return to the previous behaviour during configure
2329 with --enable-health-write.
2331 Severity : enhancement
2333 Description: 64-bit inode version
2334 Details: : Add a on-disk 64-bit inode version for ext3 to track changes made
2335 to the inode. This will be required for version-based recovery.
2340 Description: MDS fails to start if a duplicate client export is detected
2341 Details : in some rare cases it was possible for a client to connect to
2342 an MDS multiple times. Upon recovery the MDS would detect this
2343 and fail during startup. Handle this more gracefully.
2345 Severity : enhancement
2347 Description: Add -o localflock option to simulate old noflock behaviour.
2348 Details : This will achieve local-only flock/fcntl locks coherentness.
2353 Description: log_commit_thread vs filter_destroy race leads to crash
2354 Details : Take import reference before releasing llog record semaphore
2359 Description: Wrong request locking in request set processing
2360 Details : ptlrpc_check_set wrongly uses req->rq_lock for proctect add to
2361 imp_delayed_list, in this place should be used imp_lock.
2364 Frequency : when reconnecting
2366 Description: Grant leak when OSC reconnect to OST
2367 Details : When osc reconnect ost, OST(filter) should check whether it
2368 should grant more space to client by comparing fed_grant and
2369 cl_avail_grant, and return the granted space to client instead
2370 of "new granted" space, because client will call osc_init_grant
2371 to update the client grant space info.
2374 Frequency : when client reconnects to OST
2376 Description: Grant leak when OSC does a resend and replays bulk write
2377 Details : When osc reconnect to OST, OST(filter) should clear grant info of
2378 bulk write request, because the grant info while be sync between
2379 OSC and OST when reconnect, and we should ignore the grant info
2380 these of resend/replay write req.
2385 Description: Grant space more than avaiable space sometimes.
2386 Details : When then OST is about to be full, if two bulk writing from
2387 different clients came to OST. Accord the avaliable space of the
2388 OST, the first req should be permitted, and the second one
2389 should be denied by ENOSPC. But if the seconde arrived before
2390 the first one is commited. The OST might wrongly permit second
2391 writing, which will cause grant space > avaiable space.
2394 Frequency : when client is evicted
2396 Description: Grant might be wrongly erased when osc is evicted by OST
2397 Details : when the import is evicted by server, it will fork another
2398 thread ptlrpc_invalidate_import_thread to invalidate the
2399 import, where the grant will be set to 0. While the original
2400 thread will update the grant it got when connecting. So if
2401 the former happened latter, the grant will be wrongly errased
2402 because of this race.
2407 Description: Checking Stale with correct fid
2408 Details : ll_revalidate_it should uses de_inode instead of op_data.fid2
2409 to check whether it is stale, because sometimes, we want the
2410 enqueue happened anyway, and op_data.fid2 will not be initialized.
2412 Severity : enhancement
2414 Description: update patchless client
2415 Details : Add support for patchless client with 2.6.20, 2.6.21 and RHEL 5
2418 Frequency : only with 2.4 kernel
2420 Description: random memory corruption
2421 Details : size of struct ll_inode_info is to big for union inode.u and this
2422 can be cause of random memory corruption.
2427 Description: Memory leak in recovery
2428 Details : Lov_mds_md was not free in an error handler in mds_create_object.
2429 It should also check obd_fail before fsfilt_start, otherwise if
2430 fsfilt_start return -EROFS,(failover mds during mds recovery).
2431 then the req will return with repmsg->transno = 0 and rc = EROFS.
2432 and we met hit the assert LASSERT(req->rq_reqmsg->transno ==
2433 req->rq_repmsg->transno) in ptlrpc_replay_interpret. Fcc should
2434 be freed no matter whether fsfilt_commit success or not.
2437 Frequency : only with large numbers of cores on a single node
2439 Description: Prevents from taking the superblock lock in llap_from_page for
2441 Details : using LL_ORIGIN_REMOVEPAGE origin flag instead of LL_ORIGIN_UNKNOW
2442 for llap_from_page call in ll_removepage() prevents from taking
2443 the superblock lock for a soon died page.
2448 Description: Not check open intent error before release open handle
2449 Details : in some rare cases, the open intent error is not checked before
2450 release open handle, which may cause
2451 ASSERTION(open_req->rq_transno != 0), because it tries to release
2452 the failed open handle.
2457 Description: Set cat log bitmap only after create log success.
2458 Details : in some rare cases, the cat log bitmap is set too early. and it
2459 should be set only after create log success.
2463 Description: Accessing a block bevice can re-enable I/O when Lustre is
2464 tearing down a device.
2465 Details : dev_clear_rdonly(bdev) must be called in kill_bdev() instead of
2470 Description: service threads may hog cpus when there are a lot of requests
2471 Details : Insert cond_resched to give other threads a chance to use some CPU
2476 Description: the cat log was not initialized in recovery
2477 Details : When mds(mgs) do recovery, the tgt_count might be zero, so the
2478 unlink log on mds will not be initialized until mds post
2479 recovery. And also in mds post recovery, the unlink log will
2480 initialization will be done asynchronausly, so there will be race
2481 between add unlink log and unlink log initialization.
2485 Description: brw_stats were being printed incorrectly
2486 Details : brw_stats were being printed as log2 but all of them were not
2487 recorded as log2. Also remove some code duplication arising from
2488 filter_tally_{read,write}.
2492 Frequency : rare, only in recovery.
2493 Description: ASSERTION(req->rq_type != LI_POISON) failed
2494 Details : imp_lock should be held while iterating over imp_sending_list for
2495 prevent destroy request after get timeout in ptlrpc_queue_wait.
2499 Description: replay-single.sh test 52 fails
2500 Details : A lock's skiplist need to be cleanup when it being unlinked
2501 from its resource list.
2506 Description: Short directio read returns full requested size rather than
2508 Details : Direct I/O operations should return actual amount of bytes
2509 transferred rather than requested size.
2511 Severity : enhancement
2513 Description: metadata RPC reduction (e.g. for rm performance)
2514 Details : decrease the amount of synchronous RPC between clients and servers
2515 by canceling conflicing lock before the operation on the client
2516 and packing thier handles into the main operation RPC to server.
2518 Severity : enhancement
2520 Description: add #ifdef HAVE_KERNEL_CONFIG_H
2521 Details : kernels from 2.6.19 not need include linux/config.h, but add
2522 include linux/autoconf.h in commpiler command line.
2524 Severity : enhancement
2526 Description: patchless client support for 2.6.22 kernel
2527 Details : 2.6.22 has only one visble change, SLAB_CTOR_* constants is
2528 removed. In this case we need drop using os depended interface to
2529 kmem_cache and use cfs_mem_cache API.
2534 Description: fix mal-formatted messages
2535 Details : fix some mal-formatted DEBUG_REQ and LCONSOLE_ERROR_MSG messages
2539 Frequency : always in liblustre
2540 Description: wrong IS_ERR implementation in liblustre.h
2541 Details : fix IS_ERR implementation in liblustre.h for right detect errors.
2546 Description: Correct condition for output debug message.
2547 Details : inode i_nlink equal zero is not enough for output message about
2548 disk corruption, i_ctime and i_mode should be also checked.
2552 Frequency : always in patchless client
2553 Description: add configure check for truncate_complete_page
2554 Details : improve checks for exported symbols. This allow run check without
2555 sources, but with Module.symvers shipped with kernel distribution.
2556 add check for truncate_complete_page used by patchless client.
2560 Description: sanity.sh test_77h fails with "test_77h file compare failed"
2561 Details : test_77h uses a file which was messed by other test case.
2565 Description: Not Check whether lov_tgts is NULL in some lov functions
2566 Details : Checking whether lov_tgts is NULL in some functions.
2570 Description: replace obdo_alloc() with OBDO_ALLOC macro
2571 Details : nothing special is done in obdo_alloc() function, and for
2572 debugging purpose, it needs to be replaced with macros.
2576 Description: bad return value and errno from fcntl call
2577 Details : In liblustre API, errno should be a negative value if error
2582 Description: ptlrpc_check_set() LBUG
2583 Details : In case of positive reply from server and failed client bulk
2584 callback after bulk transfer shouldn't LBUG, but process this
2585 request as erroneous.
2587 Severity : enhancement
2589 Description: VFS operations stats tool.
2590 Details : Tool which collects stats by tracking value written in pid,
2591 ppid, gid and uses llstat to generate output to plot graph using
2593 Updated lustre/utils/Makefile.am
2594 Added lustre/utils/ltrack_stats.c
2596 Severity : enhancement
2598 Description: 2.6.18 server support (lustre 1.6.1)
2599 Details : Support for 2.6.18 kernels on the server side.
2604 Description: ASSERTION(imp->imp_conn_current) failed
2605 Details : an assertion failure is hit if a client node boots and attempts to
2606 mount a lustre filesystem faster than RECONNECT_INTERVAL seconds.
2609 Frequency : only for i686
2611 Description: 1.4.11 RC1 build fails for RHEL 4, i686
2612 Details : Fixed config variable for build.
2617 Description: Updated patchess for new RHEL4 kernel
2618 Details : Updated patch inode-nr_unused-2.6.9-rhel4.patch
2619 Updated patch jbd-stats-2.6.9.patch
2620 Updated patch qsnet-rhel4-2.6.patch
2621 Updated patch quota-deadlock-on-pagelock-core.patch
2622 Updated patch vfs_intent-2.6-rhel4.patch
2623 Updated patch vfs_races-2.6-rhel4.patch
2624 Updated series file 2.6-rhel4-titech.series
2625 Updated series file 2.6-rhel4.series
2626 Updated kernel config files
2631 Description: lquota slave complains LBUG when reconnecting with mds
2633 Details : quota slave depends qctxt->lqc_import to send its quota request.
2634 This pointer will be invalid if mds did failover or broke its
2635 connect to osts, which leads to LBUG.
2638 Frequency : when qunit size is too small(less than 20M)
2640 Description: write is stopped by improper -EDQUOT
2641 Details : If the master is busy and qunit size is small enough(let's say 1M),
2642 the slave can not get quota from master on time, which will lead
2643 slave to trigger a -EQUOTA to client.
2648 Description: Deadlock during metadata tests
2649 Details : in prune_dir_dentries(), shrink_dcache_parent() should not be
2650 called with the per-dentry lock held.
2653 Frequency : SLES9 only
2655 Description: Lustre patched kernel for SLES9 SP3 has NR_CPUS set to 8
2656 Details : set CONFIG_NR_CPUS to 128 instead of 8.
2658 Severity : enhancement
2660 Description: remove fs_prep_san_write operation and related patches
2661 Details : remove the ext3-san-jdike patches which are no longer useful.
2666 Description: LDISKFS-fs error (device sdc): ldiskfs_free_blocks
2667 Details : a disk corruption can cause the mballoc code to assert on a
2668 double free or other extent corruptions. Handle these with
2669 ext3_error() instead of assertions.
2672 Frequency : only with mballoc3 code and deep extent trees
2674 Description: ldiskfs_ext_search_right: bad header in inode: unexpected eh_depth
2675 Details : a wrong check of extent headers in ldiskfs_ext_search_right()
2676 can cause the filesystem to be remounted read-only.
2681 Description: lfsck built against 1.4.x cannot run against 1.6.0 lustre
2682 Details : the definition for OBD_IOC_GETNAME changed in 1.6.0. One of the
2683 few external users of this ioctl number is lfsck's call to
2684 llapi_lov_get_uuids() and this caused lfsck to fail at startup.
2685 Add the old ioctl number to the handler so both old and new
2690 Description: parallel lock callbacks
2691 Details : Instead of sending blocking and completion callbacks as separated
2692 requests, adding them to a set and sending in parallel.
2696 Description: Disable most debugging by default
2697 Details : To improve performance, disable most logging (for debug purposes)
2698 by default. VFSTRACE, RPCTRACE, and DLMTRACE are now off by
2699 default, and HA includes fewer messages.
2703 Frequency : only run on patchless client.
2704 Description: use do_facet on sanity.sh for test handling recoverables errors
2705 Details : use do_facet instead of direct use sysctl for set fail_loc on OST
2709 Description: hash tables for lists of nids, connections and uuids
2710 Details : Hash tables noticeably help when a lot of clients connect to a
2711 server, to faster identify duplicate connections or reconnects,
2712 also to faster find export to evict in manual eviction case.
2716 Description: Sometimes, when the server evict a client, and the client will
2717 not be evicted as soon as possible.
2718 Details : In enqueue req, the error was returned by intent, instead of
2719 rq_status which make ptlrpc layer not detect this error, and
2720 does not evict the client. So enqueue error should be returned
2724 Frequency : only at startup
2726 Description: Delay client connections to MDT until fist MDT->OST connect
2727 Details : If a client tried to create a new file before the MDT had
2728 connected to any OSTs, the create would return EIO. Now
2729 the client will simply block until the MDT connects to the
2730 first OST and the create can succeed.
2733 Frequency : only for SLES9
2735 Description: Routinely utilize latest Quadrics drivers in CFS releases
2736 Details : Update patch qsnet-suse-2.6.patch.
2739 Frequency : only for sles10
2741 Description: Update patches for SLES 10 SP1 kernel.
2742 Details : Update the patch vfs_intent-2.6-sles10.patch.
2747 Description: Routinely utilize latest Quadrics drivers in CFS releases
2748 Details : Update patch qsnet-rhel4-2.6.patch.
2751 Frequency : at statup only
2753 Description: mds_lov_synchronize race leads to various problems
2754 Details : simultaneous MDT->OST connections at startup can cause the
2755 sync to abort, leaving the OSC in a bad state.
2760 Description: Using wrong pointer in osc_brw_prep_request
2761 Details : Access to array[-1] can produce panic if kernel compiled with
2762 CONFIG_PAGE_ALLOC enabled
2764 Severity : enhancement
2766 Description: Async OSC create to avoid the blocking unnecessarily.
2767 Details : If a OST has no remain object, system will block on the creating
2768 when need to create a new object on this OST. Now, ways use
2769 pre-created objects when available, instead of blocking on an
2770 empty osc while others are not empty. If we must block, we block
2771 for the shortest possible period of time.
2775 Frequency : only in recovery
2776 Description: Mark OST as early accessible if his start SYNC.
2777 Details : osc_precreate return flag early accessible if oscc marked as
2778 OSCC_FLAG_SYNC_IN_PROGRESS.
2783 Description: Sometimes precreate code can triger create object on wrong ost
2784 Details : Wrong protected or not not restored variables aftre precreate loop
2785 can produce creation object on wrong ost.
2788 Frequency : oss recovery
2790 Description: llog_commit_thread cleanup should sync with llog_commit_thread
2792 Details : llog_commit_thread_count should be synced between llog_commit
2793 start and cleanup, so new llog_commit thread should not be started
2794 when llog_commit threads being stopped to avoid accessing some
2797 Severity : enhancement
2799 Description: Add printing inode info into message about error in writepage.
2801 --------------------------------------------------------------------------------
2803 2007-05-03 Cluster File Systems, Inc. <info@clusterfs.com>
2808 Frequency : on some architectures
2810 Description: 1.6 client sometimes fails to mount from a 1.4 MDT
2811 Details : Uninitialized flags sometimes cause configuration commands to
2815 Frequency : patchless clients only
2817 Description: missing __iget() symbol export
2818 Details : The __iget() symbol export is missing. To avoid the need for
2819 this on patchless clients the deathrow inode reaper is turned
2820 off, and we depend on the VM to clean up old inodes. This
2821 dependency was during via the fix for bug 12181.
2826 Description: sanity.sh fail: test_52b
2827 Details : The ll_inode_to_ext_flags() has a glitch which makes MDS return
2828 incorrect inode's flags to client.
2830 --------------------------------------------------------------------------------
2832 2007-04-19 Cluster File Systems, Inc. <info@clusterfs.com>
2834 * CONFIGURATION CHANGE. This version of Lustre WILL NOT
2835 INTEROPERATE with older versions automatically. In many cases a
2836 special upgrade step is needed. Please read the
2837 user documentation before upgrading any part of a 1.4.x system.
2838 * WARNING: Lustre configuration and startup changes are required with
2839 this release. See https://mail.clusterfs.com/wikis/lustre/MountConf
2841 * Support for kernels:
2842 2.4.21-47.0.1.EL (RHEL 3),
2843 2.6.5-7.283 (SLES 9),
2844 2.6.9-42.0.10.EL (RHEL 4),
2845 2.6.12.6 vanilla (kernel.org),
2846 2.6.16.27-0.9 (SLES10)
2847 * Client support for unpatched kernels:
2848 (see https://mail.clusterfs.com/wikis/lustre/PatchlessClient)
2849 2.6.16 - 2.6.19 vanilla (kernel.org),
2850 2.6.9-42.0.8EL (RHEL 4)
2851 * Recommended e2fsprogs version: 1.39.cfs6
2852 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
2855 Severity : enhancement
2857 Description: Async OSC create to avoid the blocking unnecessarily.
2858 Details : If a OST has no remain object, system will block on the creating
2859 when need to create a new object on this OST. Now, ways use
2860 pre-created objects when available, instead of blocking on an
2861 empty osc while others are not empty. If we must block, we block
2862 for the shortest possible period of time.
2864 Severity : enhancement
2866 Description: MountConf
2867 Details : Lustre configuration is now managed via mkfs and mount
2868 commands instead of lmc and lconf. New obd types (MGS, MGC)
2869 are added for dynamic configuration management. See
2870 https://mail.clusterfs.com/wikis/lustre/MountConf for
2873 Severity : enhancement
2875 Description: dynamic OST addition
2876 Details : OSTs can now be added to a live filesystem
2878 Severity : enhancement
2880 Description: startup order invariance
2881 Details : MDTs and OSTs can be started in any order. Clients only
2882 require the MDT to complete startup.
2884 Severity : enhancement
2886 Description: parallel, asynchronous orphan cleanup
2887 Details : orphan cleanup is now performed in separate threads for each
2888 OST, allowing parallel non-blocking operation.
2890 Severity : enhancement
2892 Description: optimized stripe assignment
2893 Details : stripe assignments are now made based on ost space available,
2894 ost previous usage, and OSS previous usage, in order to try
2895 to optimize storage space and networking resources.
2897 Severity : enhancement
2899 Description: Permanently set tunables
2900 Details : All writable /proc/fs/lustre tunables can now be permanently
2901 set on a per-server basis, at mkfs time or on a live
2904 Severity : enhancement
2906 Description: Lustre message v2
2907 Details : Add lustre message format v2.
2909 Severity : enhancement
2911 Description: client OST exclusion list
2912 Details : Clients can be started with a list of OSTs that should be
2913 declared "inactive" for known non-responsive OSTs.
2917 Description: ENOENT returned for valid filehandle during dbench.
2918 Details : Check if a directory has children when invalidating dentries
2919 associated with an inode during lock cancellation. This fixes
2920 an incorrect ENOENT sometimes seen for valid filehandles during
2921 testing with dbench.
2924 Frequency : SFS test only (otherwise harmless)
2926 Description: SPEC SFS validation failure on NFS v2 over lustre.
2927 Details : Changes the blocksize for regular files to be 2x RPC size,
2928 and not depend on stripe size.
2930 Severity : enhancement
2932 Description: fine-grained SMP locking inside DLM
2933 Details : Improve DLM performance on SMP systems by removing the single
2934 per-namespace lock and replace it with per-resource locks.
2936 Severity : enhancement
2938 Description: don't hold multiple extent locks at one time
2939 Details : To avoid client eviction during large writes, locks are not
2940 held on multiple stripes at one time or for very large writes.
2941 Otherwise, clients can block waiting for a lock on a failed OST
2942 while holding locks on other OSTs and be evicted.
2944 Severity : enhancement
2946 Description: Multiple MD RPCs in flight.
2947 Details : Further unserialise some read-only MDT RPCs - learn about intents.
2948 To avoid overly-overloading MDT, introduce a limit on number of
2949 MDT RPCs in flight for a single client and add /proc controls
2950 to adjust this limit.
2952 Severity : enhancement
2954 Description: client read/write statistics
2955 Details : Add client read/write call usage stats for performance
2956 analysis of user processes.
2957 /proc/fs/lustre/llite/*/offset_stats shows non-sequential
2958 file access. extents_stats shows chunk size distribution.
2959 extents_stats_per_process show chunk size distribution per
2962 Severity : enhancement
2964 Description: per-client statistics on server
2965 Details : Add ldlm and operations statistics for each client in
2966 /proc/fs/lustre/mds|obdfilter/*/exports/
2968 Severity : enhancement
2970 Description: improved MDT statistics
2971 Details : Add detailed MDT operations statistics in
2972 /proc/fs/lustre/mds/*/stats
2974 Severity : enhancement
2976 Description: VFS operations stats
2977 Details : Add client VFS call stats, trackable by pid, ppid, or gid
2978 /proc/fs/lustre/llite/*/stats_track_[pid|ppid|gid]
2983 Description: Fix client-side osc byte counters
2984 Details : The osc read/write byte counters in
2985 /proc/fs/lustre/osc/*/stats are now working
2988 Frequency : always as root on SLES
2990 Description: Failure of copying files with lustre special EAs.
2991 Details : Client side always return success for setxattr call for lustre
2992 special xattr (currently only "trusted.lov").
2997 Description: Refcount LNET uuids
2998 Details : The global LNET uuid list grew linearly with every startup;
2999 refcount repeated list entries instead of always adding to
3002 Severity : enhancement
3004 Description: Dynamic service threads
3005 Details : Within a small range, start extra service threads
3006 automatically when the request queue builds up.
3009 Frequency : mixed-endian client/server environments
3011 Description: mixed-endian crashes
3012 Details : The new msg_v2 system had some failures in mixed-endian
3015 Severity : enhancement
3017 Description: Easy OST removal
3018 Details : OSTs can be permanently deactivated with e.g. 'lctl
3019 conf_param lustre-OST0001.osc.active=0'
3021 Severity : enhancement
3023 Description: MGS proc entries
3024 Details : Added basic proc entries for the MGS showing what filesystems
3027 Severity : enhancement
3029 Description: provide MGS failover
3030 Details : Added config lock reacquisition after MGS server failover.
3032 Severity : enhancement
3034 Description: add Linux 2.4 support
3035 Details : Added support for RHEL 2.4.21 kernel for 1.6 servers and clients
3039 Description: a large application tries to do I/O to the same resource and dies
3040 in the middle of it.
3041 Details : Check the req->rq_arrival time after the call to
3042 ost_brw_lock_get(), but before we do anything about
3043 processing it & sending the BULK transfer request. This
3044 should help move old stale pending locks off the queue as
3045 quickly as obd_timeout.
3048 Frequency : when an incorrect nid is specified during startup
3050 Description: ptlrpc connect to non-existant node causes kernel crash
3051 Details : LNET can't be re-entered from an event callback, which
3052 happened when we expire a message after the export has been
3053 cleaned up. Instead, hand the zombie cleanup off to another
3056 Severity : enhancement
3058 Description: plain/inodebits lock performance improvement
3059 Details : Grouping plain/inodebits in granted list by their request modes
3060 and bits policy, thus improving the performance of search through
3064 Frequency : only if OST filesystem is corrupted
3066 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3067 Details : for a short time RPCs with bulk IO are in the replay list,
3068 but replay of bulk IOs is unimplemented. If the OST filesystem
3069 is corrupted due to disk cache incoherency and then replay is
3070 started it is possible to trip an assertion. Avoid putting
3071 committed RPCs into the replay list at all to avoid this issue.
3074 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3076 Description: System hang on startup
3077 Details : This bug allowed the liblustre (e.g. catamount) client to
3078 return to the app before handling all startup RPCs. This
3079 could leave the node unresponsive to lustre network traffic
3080 and manifested as a server ptllnd timeout.
3082 Severity : enhancement
3084 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3085 Details : liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT
3086 boolean to control whether to print peer debug info when a
3087 client's RPC times out.
3090 Frequency : only for kernels with patches from Lustre below 1.4.3
3092 Description: Remove old rdonly API
3093 Details : Remove old rdonly API which unused from at least lustre 1.4.3
3096 Frequency : only for devices with external journals
3098 Description: Set external device read-only also
3099 Details : During a commanded failover stop, we set the disk device
3100 read-only while the server shuts down. We now also set any
3101 external journal device read-only at the same time.
3104 Frequency : when upgrading from 1.4 while trying to change parameters
3106 Description: The wrong (new) MDC name was used when setting parameters for
3107 upgraded MDT's. Also allows changing of OSC (and MDC)
3108 parameters if --writeconf is specified at tunefs upgrade time.
3111 Frequency : when setting specific ost indicies
3113 Description: QOS code breaks on skipped indicies
3114 Details : Add checks for missing OST indicies in the QOS code, so OSTs
3115 created with --index need not be sequential.
3117 Severity : enhancement
3119 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3120 Details : The uninit_groups feature works in conjunction with the kernel
3121 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3122 up the pass1 processing of e2fsck. This is a read-only feature
3123 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3124 mount filesystems that have had this feature enabled.
3126 Severity : enhancement
3128 Description: Improve multi-block allocation algorithm to avoid fragmentation
3129 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3130 allocation locality and avoid filesystem fragmentation.
3132 ------------------------------------------------------------------------------
3134 2007-04-01 Cluster File Systems, Inc. <info@clusterfs.com>
3136 * Support for kernels:
3137 2.4.21-47.0.1.EL (RHEL 3)
3138 2.6.5-7.283 (SLES 9)
3139 2.6.9-42.0.10.EL (RHEL 4)
3140 2.6.12.6 vanilla (kernel.org)
3141 2.6.16.27-0.9 (SLES 10)
3142 * Recommended e2fsprogs version: 1.39.cfs5
3144 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3148 Frequency : occasional, depends on client load and configuration
3149 Bugzilla : 12181, 12203
3150 Description: data loss for recently-modified files
3152 Details : In some cases it is possible that recently written or created
3153 files may not be written to disk in a timely manner (this should
3154 normally be within 30s unless client IO load is very high).
3155 The problem appears as zero-length files or files that are a
3156 multiple of 1MB in size after a client crash or client eviction
3157 that are missing data at the end of the file.
3159 This problem is more likely to be hit on clients where files are
3160 repeatedly created and unlinked in the same directory, clients
3161 have a large amount of RAM, have many CPUs, the filesystem has
3162 many OSTs, the clients are rebooted frequently, and/or the files
3163 are not accessed by other nodes after being written.
3165 The presence of the problem can be detected by looking at
3166 /proc/sys/fs/inode-state. If the first number (nr_inodes) is
3167 smaller than the second (nr_unused) then dirty files will not
3168 be flushed automatically to disk. "sync; sleep 10" should be
3169 run several times on the node before unmounting it to update
3170 Lustre (this is also safe to run on nodes without this problem).
3172 There is also a related kernel bug in the RHEL4 4 2.6.9 kernel
3173 that can cause this same problem, so customers using that kernel
3174 also need to update the kernel in addition to Lustre. In order
3175 to properly fix this bug, the RHEL3 2.4.21 kernel is also updated.
3177 It is normal that files written just before a client crash (less
3178 than 30s) may not yet have been flushed to disk, even for local
3182 Frequency : frequent on thin XT3 nodes
3184 Description: UUID collision on thin XT3 Linux nodes
3185 Details : UUIDs on Compute Node Linux XT3 nodes were not generated
3186 randomly, since we relied on an insufficiently-seeded PRNG.
3191 Description: OSS hangs after "All ost request buffers busy"
3192 Details : A deadlock between quota and journal operations caused OSS
3193 hangs after printing "All ost request buffers busy."
3196 Frequency : always on liblustre builds
3198 Description: Cleanup compiler warnings on liblustre
3201 Frequency : always on liblustre builds on XT3
3203 Description: LC_CONFIG_CDEBUG don't run while build liblustre on XT3.
3207 Description: Addition of EXT3_FEATURE_RO_COMPAT_DIR_NLINKS flag for
3208 > 32000 subdirectories
3209 Details : Add EXT3_FEATURE_RO_COMPAT_DIR_NLINK flag to
3210 EXT3_FEATURE_RO_COMPAT_SUPP. This flag will be set whenever
3211 subdirectory count crosses 32000. This will aid e2fsck to
3212 correctly handle more than 32000 subdirectories.
3215 Frequency : liblustre (e.g. catamount) on a large cluster with >= 8 OSTs/OSS
3217 Description: System hang on startup
3218 Details : This bug allowed the liblustre (e.g. catamount) client to
3219 return to the app before handling all startup RPCs. This
3220 could leave the node unresponsive to lustre network traffic
3221 and manifested as a server ptllnd timeout.
3223 Severity : enhancement
3225 Description: Add "/proc/sys/lustre/debug_peer_on_timeout"
3226 (liblustre envirable: LIBLUSTRE_DEBUG_PEER_ON_TIMEOUT)
3227 boolean to control whether to print peer debug info when a
3228 client's RPC times out.
3233 Description: make O_SYNC working on 2.6 kernels
3234 Details : 2.6 kernels use different method for mark pages for write,
3235 so need add a code to lustre for O_SYNC work.
3240 Description: Failure to close file and release space on NFS
3241 Details : Put inode details into lock acquired in ll_intent_file_open.
3242 Use mdc_intent_lock in ll_intent_open to properly
3243 detect all kind of errors unhandled by mdc_enqueue.
3248 Description: proc file read during shutdown sometimes raced obd removal,
3250 Details : Add lock to prevent obd access after proc file removal.
3253 Frequency : Only for files larger than 4GB on 32-bit clients.
3255 Description: improperly doing page alignment of locks
3256 Details : Modify lustre core code to use CFS_PAGE_* defines instead of
3257 PAGE_*. Make CFS_PAGE_MASK a 64-bit mask.
3262 Description: RPCs being resent when they shouldn't be
3263 Details : Some RPCs that should not be resent are being resent. This
3264 can cause inconsistencies in the RPC state machine. Do not
3265 resend such requests.
3268 Frequency : rare, only with NFS export
3270 Description: Crash on NFS re-export node
3271 Details : under very unusual load conditions an assertion is hit in
3272 ll_intent_file_open()
3275 Frequency : only if OST filesystem is corrupted
3277 Description: client incorrectly hits assertion in ptlrpc_replay_req()
3278 Details : for a short time RPCs with bulk IO are in the replay list,
3279 but replay of bulk IOs is unimplemented. If the OST filesystem
3280 is corrupted due to disk cache incoherency and then replay is
3281 started it is possible to trip an assertion. Avoid putting
3282 committed RPCs into the replay list at all to avoid this issue.
3287 Description: large O_DIRECT requests fail under memory pressure/fragmentation
3288 Details : Large single O_DIRECT read and write calls can fail to allocate
3289 a sufficiently large buffer to process the request. In case of
3290 allocation failure the allocation is retried with a smaller
3291 buffer and broken into smaller requests.
3293 Severity : enhancement
3295 Description: Add -o localflock option to simulate old noflock behaviour.
3296 Details : This will achieve local-only flock/fcntl locks coherentness.
3301 Description: versioning check is incomplete
3302 Details : Checking the version difference of client vs. server, report
3303 error if the gap is too big.
3308 Description: add support PG_writeback bit
3309 Details : add support for PG_writeback bit for Lustre, for more carefull
3310 work with page cache in 2.6 kernel. This also fix some deadlocks
3311 and remove hack for work O_SYNC with 2.6 kernel.
3313 Severity : enhancement
3315 Description: Add uninit_groups feature to ldiskfs2 to speed up e2fsck
3316 Details : The uninit_groups feature works in conjunction with the kernel
3317 filesystem code (ldiskfs2 only) and e2fsprogs-1.39-cfs6 to speed
3318 up the pass1 processing of e2fsck. This is a read-only feature
3319 in ldiskfs2 only, so older kernels and current ldiskfs cannot
3320 mount filesystems that have had this feature enabled.
3322 Severity : enhancement
3324 Description: Improve multi-block allocation algorithm to avoid fragmentation
3325 Details : The mballoc3 code (ldiskfs2 only) adds new mechanisms to improve
3326 allocation locality and avoid filesystem fragmentation.
3328 ------------------------------------------------------------------------------
3330 2007-02-09 Cluster File Systems, Inc. <info@clusterfs.com>
3332 * Support for kernels:
3333 2.6.9-42.0.3.EL (RHEL 4)
3334 2.6.5-7.276 (SLES 9)
3335 2.4.21-47.0.1.EL (RHEL 3)
3336 2.6.12.6 vanilla (kernel.org)
3337 2.6.16.21-0.8 (SLES10)
3338 * Recommended e2fsprogs version: 1.39.cfs2-0
3340 * The backwards-compatible /proc/sys/portals symlink has been removed
3341 in this release. Before upgrading, please ensure that you change
3342 any configuration scripts or /etc/sysctl.conf files that access
3343 /proc/sys/portals/* or sysctl portals.* to use the corresponding
3344 entry in /proc/sys/lnet or sysctl lnet.*. This change can be made
3345 in advance of the upgrade on any system running Lustre 1.4.6 or
3346 newer, since /proc/sys/lnet was added in that version.
3347 * Note that reiserfs quotas are disabled on SLES 10 in this kernel
3351 Frequency : only when quota is used
3353 Description: avoid scanning export list for quota master
3354 Details : Change the algorithms to avoid scanning export list in order
3355 to improve the efficiency.
3358 Frequency : MDS failover only, very rarely
3360 Description: "went back in time" messages on mds failover
3361 Details : The greatest transno may be lost when the current operation
3362 finishes with an error (transno==0) and the client's last_rcvd
3363 record is over-written. Save the greatest transno in the
3364 mds_last_transno for this case.
3367 Frequency : always for specific kernels and striping counts
3369 Description: client may get "Matching packet too big" without ACL support
3370 Details : Clients compiled without CONFIG_FS_POSIX_ACL get an error message
3371 when trying to access files in certain configurations. The
3372 clients should in fact be denied when mounting because they do
3373 not understand ACLs.
3376 Frequency : Cray XT3 with more than 4000 clients and multiple jobs
3378 Description: many clients connecting with IO in progress causes connect timeouts
3379 Details : Avoid synchronous journal commits to avoid delays caused by many
3380 clients connecting/disconnecting when bulk IO is in progress.
3381 Queue liblustre connect requests on OST_REQUEST_PORTAL instead of
3382 OST_IO_PORTAL to avoid delays behind potentially many pending
3386 Frequency : occasionally with multiple writers to a single file
3388 Description: shared writes to file may result in wrong size reported by stat()
3389 Details : Allow growing of kms when extent lock is cancelled
3392 Frequency : always with random mmap IO to multi-striped file
3394 Description: mmap write might be lost if we are writing to a 'hole' in stripe
3395 Details : Only if the hole is at the end of OST object so that kms is too
3396 small. Fix is to increase kms accordingly in ll_nopage.
3399 Frequency : rare, only if OST filesystem is inconsistent with MDS filesystem
3401 Description: writes to a missing object would leak memory on the OST
3402 Details : If there is an inconsistency between the MDS and OST filesystems,
3403 such that the MDS references an object that doesn't exist, writes
3404 to that object will leak memory due to incorrect cleanup in the
3405 error handling path, eventually running out of memory on the OST.
3410 Description: Creating too long symlink causes lustre errors
3411 Details : Check symlink and name lengths before sending requests to MDS.
3414 Frequency : only if flock is enabled (not on by default)
3416 Description: posix locks not released on fd closure on 2.6.9+
3417 Details : We failed to add posix locks to list of inode locks on 2.6.9+
3418 kernels, this caused such locks not to be released on fd close and
3419 then assertions on fs unmount about still used locks.
3422 Frequency : MDS failover only, very rarely
3424 Description: clients may get ASSERTION(granted_lock != NULL)
3425 Details : When request was taking a long time, and a client was resending
3426 a getattr by name lock request. The were multiple lock requests
3427 with the same client lock handle and
3428 mds_getattr_name->fixup_handle_for_resent_request found one of the
3429 lock handles but later failed with ASSERTION(granted_lock != NULL).
3434 Description: handle->h_buffer_credits > 0, assertion failure
3435 Details : h_buffer_credits is zero after truncate, causing assertion
3436 failure. This patch extends the transaction or creates a new
3440 Frequency : NFS re-export or patchless client
3441 Bugzilla : 11179, 10796
3442 Description: Crash on NFS re-export node (__d_move)
3443 Details : We do not want to hash the dentry if we don't have a lock.
3444 But if this dentry is later used in d_move, we'd hit uninitialised
3445 list head d_hash, so we just do this to init d_hash field but
3446 leave dentry unhashed.
3449 Frequency : NFS re-export or patchless client
3451 Description: NFS exports has problem with symbolic link
3452 Details : lustre client didn't properly install dentry when re-exported
3453 to NFS or running patchless client.
3456 Frequency : NFS re-export or patchless client
3458 Description: Various nfs/patchless fixes.
3459 Details : fixes reuse disconected alias for lookup process - this fixes
3460 warning "find_exported_dentry: npd != pd",
3461 fix permission error with open files at nfs.
3462 fix apply umask when do revalidate.
3465 Frequency : occasional
3467 Description: Crash on NFS re-export node
3468 Details : calling clear_page() on the wrong pointer triggered oops in
3469 generic_mapping_read().
3472 Frequency : rarely, using O_DIRECT IO
3474 Description: unaligned directio crashes client with LASSERT
3475 Details : check for unaligned buffers before trying any requests.
3478 Frequency : rarely, using CFS RAID5 patches in non-standard kernel series
3480 Description: stale data returned from RAID cache
3481 Details : If only a small amount of IO is done to the RAID device before
3482 reading it again it is possible to get stale data from the RAID
3483 cache instead of reading it from disk.
3486 Frequency : always for sles10 kernel
3488 Description: sles10 support
3489 Details : ll_follow_link: compile fixes and using of nd_set_link
3490 under newer kernels.
3493 Frequency : depends on arch, kernel and compiler version, always on sles10
3496 Description: recursive or deep enough symlinks cause stack overflow
3497 Details : getting rid of large stack-allocated variable in
3501 Frequency : depends on hardware
3503 Description: lustre write performance loss in the SLES10 kernel
3504 Details : the performance loss is caused by using of write barriers in the
3505 ext3 code. The SLES10 kernel turns barrier support on by
3506 default. The fix is to undo that change for ldiskfs.
3508 ------------------------------------------------------------------------------
3510 2006-12-09 Cluster File Systems, Inc. <info@clusterfs.com>
3512 * Support for kernels:
3513 2.6.9-42.0.3EL (RHEL 4)
3514 2.6.5-7.276 (SLES 9)
3515 2.4.21-47.0.1.EL (RHEL 3)
3516 2.6.12.6 vanilla (kernel.org)
3520 Frequency : quota enabled and large files being deleted
3522 Description: releasing more than 4GB of quota at once hangs OST
3523 Details : If a user deletes more than 4GB of files on a single OST it
3524 will cause the OST to spin in an infinite loop. Release
3525 quota in < 4GB chunks, or use a 64-bit value for 1.4.7.1+.
3530 Description: statfs data retrieved from /proc may be stale or zero
3531 Details : When reading per-device statfs data from /proc, in the
3532 {kbytes,files}_{total,free,avail} files, it may appear
3533 as zero or be out of date.
3536 Frequency : systems with MD RAID1 external journal devices
3538 Description: lconf's call to blkid is confused by RAID1 journal devices
3539 Details : Use the "blkid -l" flag to locate the MD RAID device instead
3540 of returning all block devices that match the journal UUID.
3543 Frequency : always, for aggregate stripe size over 4GB
3545 Description: "lfs setstripe" fails assertion when setting 4GB+ stripe width
3546 Details : Using "lfs setstripe" to set stripe size * stripe count over 4GB
3547 will fail the kernel with "ASSERTION(lsm->lsm_xfersize != 0)"
3550 Frequency : always if "lfs find" used on a local file/directory
3552 Description: "lfs find" segfaults if used on a local file/directory
3553 Details : The case where a directory component was not specified wasn't
3554 handled correctly. Handle this properly.
3557 Frequency : always on ppc64
3559 Description: the write to an ext3 filesystem mounted with mballoc got stuck
3560 Details : ext3_mb_generate_buddy() uses find_next_bit() which does not
3561 perform endianness conversion.
3564 Frequency : rarely (truncate to non-zero file size after write under load)
3565 Bugzilla : 10730, 10687
3566 Description: Files padded with zeros to next 4K multiple
3567 Details : With filesystems mounted using the "extents" option (2.6 kernels)
3568 it is possible that files that are truncated to a non-zero size
3569 immediately after being written are filled with zero bytes beyond
3570 the truncated size. No file data is lost.
3572 Severity : enhancement
3574 Description: Allow recovery/failover for liblustre clients.
3575 Details : liblustre clients were unaware of failover configurations until
3578 Severity : enhancement
3580 Description: user file locks should fail when not mounting with flock option
3581 Details : Set up an error-returning stub in ll_file_operations.lock field
3582 to prevent incorrect behaviour when client is mounted without
3583 flock option. Also, set up properly f_op->flock field for
3587 Frequency : always on ia64
3589 Description: "lfs df" loops on printing out MDS statfs information
3590 Details : The obd_ioctl_data was not initialized and in some systems
3591 this caused a failure during the ioctl that did not return
3592 an error. Initialize the struct and return an error on failure.
3595 Frequency : SLES 9 only
3597 Description: Error of copying files with lustre special EAs as root
3598 Details : Client side always return success for setxattr call for lustre
3599 special xattr (currently only "trusted.lov").
3602 Frequency : rarely on clusters with both ia64+i386 clients
3604 Description: ia64+i686 clients doing shared IO on the same file may LBUG
3605 Details : In rare cases when both ia64+i686 (or other mixed-PAGE_SIZE)
3606 clients are doing concurrent writes to the same file it is
3607 possible that the ia64 clients may LASSERT because the OST
3608 extent locks are not PAGE_SIZE aligned. Ensure that grown
3609 locks are always aligned on the request boundary.
3612 Frequency : specific use, occasional
3614 Description: Overwriting in use executable truncates on-disk binary image
3615 Details : If one node attempts to overwrite an executable in use by
3616 another node, we now correctly return ETXTBSY instead of
3617 truncating the file.
3619 Severity : enhancement
3621 Description: Async OSC create to avoid the blocking unnecessarily.
3622 Details : If a OST has no remain object, system will block on the creating
3623 when need to create a new object on this OST. Now, ways use
3624 pre-created objects when available, instead of blocking on an
3625 empty osc while others are not empty. If we must block, we block
3626 for the shortest possible period of time.
3631 Description: chmod on Lustre root is propagated to other clients
3632 Details : Re-validate root's dentry in ll_lookup_it to avoid having it
3633 invalid by the follow_mount time.
3636 Frequency : liblustre clients only
3638 Description: Race in 'instant cancel' lock handling could lead to such locks
3639 never to be granted in case of SMP MDS
3640 Details : Do not destroy not yet granted but cbpending locks in
3644 Frequency : replay/resend of open
3646 Description: non null lock assetion failure in mds_intent_policy
3647 Details : Trying to replay/resend lockless open requests resulted in
3648 mds_open() returning 0 with no lock. Now it sets a flag if
3649 a lock is going to be returned.
3651 Severity : enhancement
3653 Description: Checksum enhancements
3654 Details : New checksum enhancements allow for resending RPCs that failed
3657 Severity : enhancement
3659 Description: Tunables on number of dirty pages in cacche
3660 Details : Allow to set limit on number of dirty pages cached.
3665 Description: client crash on unmount - lock still has references
3666 Details : In some error handling cases it was possible to leak a lock
3667 reference on a client while accessing a file. This was not
3668 harmful to the client during operation, but would cause the
3669 client to crash when the filesystem is unmounted.
3672 Frequency : specific case, rare
3674 Description: ETXTBSY on mds though file not in use by client
3675 Details : ETXTBSY is no longer incorrectly returned when attempting to
3676 chmod or chown a directory that the user previously tried to
3677 execute or a currently-executing binary.
3680 Frequency : extremely rare except on liblustre-based clients
3682 Description: Lustre space not freed when files are deleted
3683 Details : Clean up open-unlinked files after client eviction. Previously
3684 the unlink was skipped and the files remained as orphans.
3689 Description: OST failure "would be an LBUG" in waiting_locks_callback()
3690 Details : In some cases it was possible to send a blocking callback to
3691 a client doing a glimpse, even though that client didn't get
3692 a lock granted. When the glimpse lock is cancelled on the OST
3693 the freed lock is left on the waiting list and corrupted the list.
3696 Frequency : all core dumps
3698 Description: Broke core dumps to lustre
3699 Details : Negative dentry may be unhashed if parent does not have UPDATE
3700 lock, but some callers, e.g. do_coredump, expect dentry to be
3701 hashed after successful create, hash it in ll_create_it.
3703 ------------------------------------------------------------------------------
3705 2006-09-13 Cluster File Systems, Inc. <info@clusterfs.com>
3707 * Support for kernels:
3708 2.6.9-42.0.2.EL (RHEL 4)
3709 2.6.5-7.276 (SLES 9)
3710 2.4.21-40.EL (RHEL 3)
3711 2.6.12.6 vanilla (kernel.org)
3715 Frequency : always on RHEL 3
3717 Description: Number of open files grows over time
3718 Details : The number of open files grows over time, whether or not
3719 Lustre is started. This was due to a filp leak introduced
3720 by one of our kernel patches.
3722 ------------------------------------------------------------------------------
3724 2006-08-20 Cluster File Systems, Inc. <info@clusterfs.com>
3726 * Support for kernels:
3727 2.6.9-42.EL (RHEL 4)
3728 2.6.5-7.267 (SLES 9)
3729 2.4.21-40.EL (RHEL 3)
3730 2.6.12.6 vanilla (kernel.org)
3735 Bugzilla : 5719, 9635, 9792, 9684
3736 Description: OST (or MDS) trips assertions in (re)connection under heavy load
3737 Details : If a server is under heavy load and cannot reply to new
3738 connection requests before the client resends the (re)connect,
3739 the connection handling code can behave badly if two service
3740 threads are concurrently handing separate (re)connections from
3741 the same client. Add better locking to the connection handling
3742 code, and ensure that only a single connection will be processed
3743 for a given client UUID, even if the lock is dropped.
3745 Severity : enhancement
3747 Description: add TCP zero-copy support to kernel
3748 Details : Add support to the kernel TCP stack to allow zero-copy bulk
3749 sends if the hardware supports scatter-gather and checksumming.
3750 This allows socklnd to do client-write and server-read more
3751 efficiently and reduce CPU utilization from skbuf copying.
3754 Frequency : only if NFS exporting from client
3756 Description: NULL pointer deref in ll_iocontrol() if chattr mknod file
3757 Details : If setting attributes on a file created under NFS that had
3758 never been opened it would be possible to oops the client
3759 if the file had no objects.
3763 Bugzilla : 9326, 10402, 10897
3764 Description: client crash in ptlrpcd_wake() thread when sending async RPC
3765 Details : It is possible that ptlrpcd_wake() dereferences a freed async
3766 RPC. In rare cases the ptlrpcd thread alread processed the RPC
3767 before ptlrpcd_wake() was called and the request was freed.
3770 Frequency : always for liblustre
3772 Description: liblustre client does MDS+OSTs setattr RPC for each write
3773 Details : When doing a write from a liblustre client, the client
3774 incorrectly issued an RPC to the MDS and each OST the file was
3775 striped over in order to update the timestamps. When writing
3776 with small chunks and many clients this could overwhelm the MDS
3777 with RPCs. In all cases it would slow down the write because
3778 these RPCs are unnecessary.
3780 Severity : enhancement
3782 Description: allow number of MDS service threads to be changed at module load
3783 Details : It is now possible to change the number of MDS service threads
3784 running. Adding "options mds mds_num_threads={N}" to the MDS's
3785 /etc/modprobe.conf will set the number of threads for the next
3786 time Lustre is restarted (assuming the "mds" module is also
3787 reloaded at that time). The default number of threads will
3788 stay the same, 32 for most systems.
3793 Description: OST crash if filesystem is unformatted or corrupt
3794 Details : If an OST is started on a device that has never been formatted
3795 or if the filesystem is corrupt and cannot even mount then the
3796 error handling cleanup routines would dereference a NULL pointer.
3801 Description: NULL pointer deref in llap_from_page.
3802 Details : get_cache_page_nowait can return a page with NULL (or otherwise
3803 incorrect) mapping if the page was truncated/reclaimed while it was
3804 searched for. Check for this condition and skip such pages when
3805 doing readahead. Introduce extra check to llap_from_page() to
3806 verify page->mapping->host is non-NULL (so page is not anonymous).
3809 Frequency : Sometimes when using sys_sendfile
3811 Description: "page not covered by a lock" warnings from ll_readpage
3812 Details : sendfile called ll_readpage without right page locks present.
3813 Now we introduced ll_file_sendfile that does necessary locking
3814 around call to generic_file_sendfile() much like we do in
3818 Frequency : with certain MDS communication failures at client mount time
3820 Description: NULL pointer deref after failed client mount
3821 Details : a client connection request may delayed by the network layer
3822 and not be sent until after the PTLRPC layer has timed out the
3823 request. If the client fails the mount immediately it will try
3824 to clean up before the network times out the request. Add a
3825 reference from the request import to the obd device and delay
3826 the cleanup until the network drops the request.
3829 Frequency : occasionally during client (re)connect
3831 Description: assertion failure during client (re)connect
3832 Details : processing a client connection request may be delayed by the
3833 client or server longer than the client connect timeout. This
3834 causes the client to resend the connection request. If the
3835 original connection request is replied in this interval, the
3836 client may trip an assertion failure in ptlrpc_connect_interpret()
3837 which thought it would be the only running connect process.
3840 Frequency : only with obd_echo servers and clients that are rebooted
3842 Description: kernel BUG accessing uninitialized data structure
3843 Details : When running an obd_echo server it did not start the ping_evictor
3844 thread, and when a client was evicted an uninitialized data
3845 structure was accessed. Start the ping_evictor in the RPC
3846 service startup instead of the OBD startup.
3848 Severity : enhancement
3849 Bugzilla : 10193 (patchless)
3850 Description: Remove dependency on various unexported kernel interfaces.
3851 Details : No longer need reparent_to_init, exit_mm, exit_files,
3852 sock_getsockopt, filemap_populate, FMODE_EXEC, put_filp.
3855 Frequency : rare (only users of deprecated and unsupported LDAP config)
3857 Description: write_conf for zeroconf mount queried LDAP incorrectly for client
3858 Details : LDAP apparently contains 'lustreName' attributes instead of
3859 'name'. A simple remapping of the name is sufficient.
3862 Frequency : rare (only with non-default dump_on_timeout debug enabled)
3864 Description: waiting_locks_callback trips kernel BUG if client is evicted
3865 Details : Running with the dump_on_timeout debug flag turned on makes
3866 it possible that the waiting_locks_callback() can try to dump
3867 the Lustre kernel debug logs from an interrupt handler. Defer
3868 this log dumping to the expired_lock_main() thread.
3870 Severity : enhancement
3872 Description: Support NFS exporting on 2.6 kernels.
3873 Details : Implement non-rawops metadata methods for NFS server to use without
3874 changing NFS server code.
3877 Frequency : very rare (synthetic metadata workload only)
3879 Description: two racing renames might cause an MDS thread to deadlock
3880 Details : Running the "racer" program may cause one MDS thread to rename
3881 a file from being the source of a rename to being the target of
3882 a rename at exactly the same time that another thread is doing
3883 so, and the second thread has already enqueued these locks after
3884 doing a lookup of the target and is trying to relock them in
3885 order. Ensure that we don't try to re-lock the same resource.
3888 Frequency : only very large systems with liblustre clients
3890 Description: slow eviction of liblustre clients with the "evict_by_nid" RPC
3891 Details : Use asynchronous set_info RPCs to send the "evict_by_nid" to
3892 all OSTs in parallel. This allows the eviction of stale liblustre
3893 clients to proceed much faster than if they were done in series,
3894 and also offers similar improvements for other set_info RPCs.
3899 Description: excessive CPU usage during initial read phase on client
3900 Details : During the initial read phase on a client, it would agressively
3901 retry readahead on the file, consuming too much CPU and impacting
3902 performance (since 1.4.5.8). Improve the readahead algorithm
3903 to avoid this, and also improve some other common cases (read
3904 of small files in particular, where "small" is files smaller than
3905 /proc/fs/lustre/llite/*/max_read_ahead_whole_mb, 2MB by default).
3910 Description: MDS crash when receiving packet with unknown intent.
3911 Details : Do not LBUG in unknown intent case, just return -EFAULT
3913 Severity : enhancement
3914 Bugzilla : 9293, 9385
3915 Description: MDS RPCs are serialised on client. This is unnecessary for some.
3916 Details : Do not serialize getattr (non-intent version) and statfs.
3919 Frequency : occasional, when OST network is overloaded/intermittent
3921 Description: client evicted by OST after bulk IO timeout
3922 Details : If a client sends a bulk IO request (read or write) the OST
3923 may evict the client if it is unresposive to its data GET/PUT
3924 request. This is incorrect if the network is overloaded (takes
3925 too long to transfer the RPC data) or dropped the OST GET/PUT
3926 request. There is no need to evict the client at all, since
3927 the pinger and/or lock callbacks will handle this, and the
3928 client can restart the bulk request.
3931 Frequency : Always when mmapping file with no objects
3933 Description: client crashes when mmapping file with no objects
3934 Details : Check that we actually have objects in a file before doing any
3935 operations on objects in ll_vm_open, ll_vm_close and
3941 Description: Request leak when working with deleted CWD
3942 Details : Introduce advanced request refcount tracking for requests
3943 referenced from lustre intent.
3945 Severity : Enhancement
3947 Description: Cache open file handles on client.
3948 Details : MDS now will return special lock along with openhandle, if
3949 requested and client is allowed to hold openhandle, even if unused,
3950 until such a lock is revoked. Helps NFS a lot, since NFS is opening
3951 closing files for every read/write openration.
3953 Severity : Enhancement
3955 Description: Cache open negative dentries on client when possible.
3956 Details : Guard negative dentries with UPDATE lock on parent dir, drop
3957 negative dentries on lock revocation.
3962 Description: Remounting a client read-only wasn't possible with a zconf mount
3963 Details : It wasn't possible to remount a client read-only with llmount.
3965 Severity : enhancement
3966 Description: Include MPICH 1.2.6 Lustre ADIO interface patch
3967 Details : In lustre/contrib/ or /usr/share/lustre in RPM a patch for
3968 MPICH is included to add Lustre-specific ADIO interfaces.
3969 This is based closely on the UFS ADIO layer and only differs
3970 in file creation, in order to allow the OST striping to be set.
3971 This is user-contributed code and not supported by CFS.
3976 Description: extended inode attributes (immutable, append-only) work improperly
3977 when 2.4 and 2.6 kernels are used on client/server or vice versa
3978 Details : Introduce kernel-independent values for these flags.
3980 Severity : enhancement
3983 Description: Allow fractional MB tunings for lustre in /proc/ filesystem.
3984 Details : Many of the /proc/ tunables can only be tuned at a megabyte
3985 granularity. Now, Fractional MB granularity is be supported,
3986 this is very useful for low memory system.
3988 Severity : enhancement
3990 Description: Getattr by fid
3991 Details : Getting a file attributes by its fid, obtaining UPDATE|LOOKUP
3992 locks, avoids extra getattr rpc requests to MDS, allows '/' to
3993 have locks and avoids getattr rpc requests for it on every stat.
3996 Frequency : Always, for filesystems larger than 2TB
3998 Description: ldiskfs crash at mount for filesystem larger than 2TB with mballoc
3999 Details : Kenrel kmalloc limits allocations to 128kB and this prevents
4000 filesystems larger than 2TB to be mounted with mballoc enabled.
4003 Frequency : Always, for 32-bit kernel without CONFIG_LBD and filesystem > 2TB
4005 Description: filesystem corruption for non-standard kernels and very large OSTs
4006 Details : If a 32-bit kernel is compiled without CONFIG_LBD enabled and a
4007 filesystems larger than 2TB is mounted then the kernel will
4008 silently corrupt the start of the filesystem. CONFIG_LBD is
4009 enabled for all CFS-supported kernels, but the possibility of
4010 this happening with a modified kernel config exists.
4012 Severity : enhancement
4014 Description: add client O_DIRECT support for 2.6 kernels
4015 Details : It is now possible to do O_DIRECT reads and writes to files
4016 in the Lustre client mountpoint on 2.6 kernel clients.
4018 Severity : enhancement
4020 Description: parallel glimpse, setattr, statfs, punch, destroy requests
4021 Details : Sends glimpse, setattr, statfs, punch, destroy requests to OSTs in
4022 parallel, not waiting for response from every OST before sending
4023 a rpc to the next OST.
4028 Description: setattr vs write race when updating file timestamps
4029 Details : Client processes that update a file timestamp into the past
4030 right after writing to the file (e.g. tar) it is possible that
4031 the updated file modification time can be reset to the current
4032 time due to a race between processing the setattr and write RPC.
4034 Severity : enhancement
4036 Description: Bring 'lfs find' closer in line with regular Linux find.
4037 Details : lfs find util supports -atime, -mtime, -ctime, -maxdepth, -print,
4038 -print0 options and obtains all the needed info through the lustre
4041 Severity : enhancement
4043 Description: support up to 1024 configured devices on one node
4044 Details : change obd_dev array from statically allocated to dynamically
4045 allocated structs as they are first used to reduce memory usage
4050 Description: Flush dirty partially truncated pages during truncate
4051 Details : Immediatelly flush partially truncated pages in filter_setattr,
4052 this way we completely avoid having any pages in page cache on OST
4053 and can retire ugly workarounds during writes to flush such pages.
4058 Description: i_sem vs transaction deadlock in mds_obd_destroy during unlink.
4059 Details : protect inode from truncation within vfs_unlink() context
4060 just take a reference before calling vfs_unlink() and release it
4061 when parent's i_sem is free.
4064 Frequency : always, if extents are used on OSTs
4066 Description: index ei_leaf_hi (48-bit extension) is not zeroed in extent index
4067 Details : OSTs using the extents format would not zero the high 16 bits of
4068 the index physical block number. This is not a problem for any
4069 OST filesystems smaller than 16TB, and no kernels support ext3
4070 filesystems larger than 16TB yet. This is fixed in 1.4.7 (all
4071 new/modified files) and can be fixed for existing filesystems
4072 with e2fsprogs-1.39-cfs1.
4077 Description: import connection selection may be incorrect if timer wraps
4078 Details : Using a 32-bit jiffies timer with HZ=1000 may cause backup
4079 import connections to be ignored if the 32-bit jiffies counter
4080 wraps. Use a 64-bit jiffies counter.
4083 Frequency : during server recovery
4085 Description: crash after server is denying duplicate export
4086 Details : If clients are resending connect requests to the server, the
4087 server refuses to allow a client to connect multiple times.
4088 Fixed a bug in the handling of this case.
4091 Frequency : very large clusters immediately after boot
4093 Description: LNET request buffers exhausted under heavy short-term load
4094 Details : If a large number of client requests are generated on a service
4095 that has previously never seen so many requests it is possible
4096 that the request buffer growth cannot keep up with the spike in
4097 demand. Instead of dropping incoming requests, they are held in
4098 the LND until the RPC service can accept more requests.
4101 Frequency : Sometimes during replay
4103 Description: Assertion failure in ll_local_open after replay.
4104 Details : If replay happened on an open request reply before we were able
4105 to set replay handler, reply will become not swabbed tripping the
4106 assertion in ll_local_open. Now we set the handler right after
4107 recognising of open request
4110 Frequency : very rare
4112 Description: kernel reports "badness in vsnprintf"
4113 Details : Reading from the "recovery_status" /proc file in small chunks
4114 may cause a negative length in lprocfs_obd_rd_recovery_status()
4115 call to vsnprintf() (which is otherwise harmless). Exit early
4116 if there is no more space in the output buffer.
4118 Severity : enhancement
4120 Description: clear OBD RPC statistics by writing to them
4121 Details : It is now possible to clear the OBD RPC statistics by writing
4122 to the "stats" file.
4127 Description: Client mtime is not the same on different clients after utimes
4128 Details : In some cases, the client was using the utimes() syscall on
4129 a file cached on another node. The clients now validate the
4130 ctime from the MDS + OSTs to determine which one is right.
4135 Description: Inability to activate failout mode
4136 Details : lconf script incorrectly assumed that in python string's numeric
4137 value is used in comparisons.
4140 Frequency : always with multiple stripes per file
4142 Description: Inefficient object allocation for mutli-stripe files
4143 Details : When selecting which OSTs to stripe files over, for files with
4144 a stripe count that divides evenly into the number of OSTs,
4145 the MDS is always picking the same starting OST for each file.
4146 Return the OST selection heuristic to the original design.
4151 Description: mount failures may take full timeout to return an error
4152 Details : Under some heavy load conditions it is possible that a
4153 failed mount can wait for the full obd_timeout interval,
4154 possibly several minutes, before reporting an error.
4155 Instead return an error as soon as the status is known.
4157 ------------------------------------------------------------------------------
4159 2006-02-14 Cluster File Systems, Inc. <info@clusterfs.com>
4161 * WIRE PROTOCOL CHANGE. This version of Lustre networking WILL NOT
4162 INTEROPERATE with older versions automatically. Please read the
4163 user documentation before upgrading any part of a live system.
4164 * WARNING: Lustre networking configuration changes are required with
4165 this release. See https://bugzilla.clusterfs.com/show_bug.cgi?id=10052
4168 * Support for kernels:
4169 2.6.9-22.0.2.EL (RHEL 4)
4170 2.6.5-7.244 (SLES 9)
4171 2.6.12.6 vanilla (kernel.org)
4174 Severity : enhancement
4175 Bugzilla : 7981/8208
4176 Description: Introduced Lustre Networking (LNET)
4177 Details : LNET is new networking infrastructure for Lustre, it includes
4178 a reorganized network configuration mode (see the user
4179 documentation for full details) as well as support for routing
4180 between different network fabrics. Lustre Networking Devices
4181 (LNDs) for the supported network fabrics have also been
4182 created for this new infrastructure.
4184 Severity : enhancement
4185 Description: Introduced Access control lists
4186 Details : clients can set ACLs on files and directories in order to have
4187 more fine-grained permissions than the standard Unix UGO+RWX.
4188 The MDS must be started with the "-o acl" mount option.
4190 Severity : enhancement
4191 Description: Introduced filesystem quotas
4192 Details : Administrators may now establish per-user quotas on the
4195 Severity : enhancement
4197 Description: Configuration change for the XT3
4198 The PTLLND is now used to run Lustre over Portals on the XT3
4199 The configure option(s) --with-cray-portals are no longer used.
4200 Rather --with-portals=<path-to-portals-includes> is used to
4201 enable building on the XT3. In addition to enable XT3 specific
4202 features the option --enable-cray-xt3 must be used.
4207 Description: Running on many-way SMP OSTs can trigger oops in llcd_send()
4208 Details : A race between allocating a new llcd and re-getting the llcd_lock
4209 allowed another thread to grab newly-allocated llcd.
4211 Severity : enhancement
4213 Description: 2.6 OST async journal commit and locking fix to improve performance
4214 Details : The filter_direct_io()+filter_commitrw_write() journal commits for
4215 2.6 kernels are now async as they already were in 2.4 kernels so
4216 that they can commit concurrently with the network bulk transfer.
4217 For block-allocated files the filter allocation semaphore is held
4218 to avoid filesystem fragmentation during allocation. BKL lock
4219 removed for 2.6 xattr operations where it is no longer needed.
4224 Description: lconf incorrectly determined whether two IP networks could talk
4225 Details : In some more complicated routing and multiple-network
4226 configurations, lconf will avoid trying to make a network
4227 connection to a disjoint part of the IP space. It was doing the
4228 math incorrectly for one set of cases.
4233 Description: Fix for potential infinite loop processing records in an llog.
4234 Details : If an llog record is corrupted/zeroed, it is possible to loop
4235 forever in llog_process(). Validate the llog record length
4236 and skip the remainder of the block on error.
4239 Frequency : occasional (liblustre only)
4241 Description: liblustre could not open files whose last component is a symlink
4242 Details : sysio_path_walk() would incorrectly pass the open intent to
4243 intermediate path components.
4246 Frequency : rare (liblustre only with non-standard tuning)
4247 Bugzilla : 7201 (7350)
4248 Description: Tuning the MDC DLM LRU size to zero triggers client LASSERT
4249 Details : llu_lookup_finish_locks() tries to set lock data on a lock
4250 after it has been released, only do this for referenced locks
4252 Severity : enhancement
4254 Description: specifying an (invalid) directory default stripe_size of -1
4255 would reset the directory default striping
4256 Details : stripe_size -1 was used internally to signal directory stripe
4257 removal, now use "all default" to signal dir stripe removal
4258 as a directory striping of "all default" is not useful
4261 Frequency : common for large clusters running liblustre clients
4263 Description: doing an ls when liblustre clients are running is slow
4264 Details : sending a glimpse AST to a liblustre client waits for every AST
4265 to time out, as liblustre clients will not respond. Since they
4266 cannot cache data we refresh the OST lock LVB from disk instead.
4268 Severity : enhancement
4270 Description: doing an ls at the same time as file IO can be slow
4271 Details : enqueue and other "small" requests can be blocked behind many
4272 large IO requests. Create a new OST IO portal for non-IO
4273 requests so they can be processed faster.
4276 Frequency : rare (only HPUX clients mounting unsupported re-exported NFS vol)
4278 Description: an HPUX NFS client would get -EACCESS when ftruncate()ing a newly
4279 created file with mode 000
4280 Details : the Linux NFS server relies on an MDS_OPEN_OWNEROVERRIDE hack to
4281 allow an ftruncate() as a non-root user to a file with mode 000.
4282 Lustre now respects this flag to disable mode checks when
4283 truncating a file owned by the user
4286 Frequency : liblustre-only, when liblustre client dies unexpectedly or becomes
4289 Description: Revoking locks from clients that went dead or catatonic might take
4291 Details : New lock flags FL_CANCEL_ON_BLOCK used by liblustre makes
4292 cancellation of such locks instant on servers without waiting for
4293 any reply from clients. Clients drops these locks when cancel
4294 notification from server is received without replying.
4297 Frequency : liblustre-only, when liblustre client dies or becomes busy
4299 Description: Doing ls on Linux clients can take a long time with active
4301 Details : Liblustre client cannot handle ASTs in timely manner, so avoid
4302 granting such locks to it in the first place if possible. Locks
4303 are taken by proxy on the OST during the read or write and
4304 dropped immediately afterward. Add connect flags handling, do
4305 not grant locks to liblustre clients for glimpse ASTs.
4307 Severity : enhancement
4309 Description: Improve read-ahead algorithm to avoid excessive IO for random reads
4310 Details : Existing read-ahead algorithm is tuned for the case of streamlined
4311 sequential reads and behaves badly with applications doing random
4312 reads. Improve it by reading ahead at least read region, and
4313 avoiding excessive large RPC for small reads.
4315 Severity : enhancement
4317 Description: Creating more than 1000 files for a single job may cause a load
4318 imbalance on the OSTs if there are also a large number of OSTs.
4319 Details : qos_prep_create() uses an OST index reseed value that is an
4320 even multiple of the number of available OSTs so that if the
4321 reseed happens in the middle of the object allocation it will
4322 still utilize the OSTs as uniformly as possible.
4327 Description: OST or MDS may oops in ping_evictor_main()
4328 Details : ping_evictor_main() drops obd_dev_lock if deleting a stale export
4329 but doesn't restart at beginning of obd_exports_timed list
4332 Severity : enhancement
4334 Description: improve by-nid export eviction on the MDS and OST
4335 Details : allow multiple exports with the same NID to be evicted at one
4336 time without re-searching the exports list.
4339 Frequency : rare, only with supplementary groups enabled on SMP 2.6 kernels
4341 Description: MDS may oops in groups_free()
4342 Details : in rare race conditions a newly allocated group_info struct is
4343 freed again, and this can be NULL. The 2.4 compatibility code
4344 for groups_free() checked for a NULL pointer, but 2.6 did not.
4347 Frequency : common for liblustre clients doing little filesystem IO
4348 Bugzilla : 9352, 7313
4349 Description: server may evict liblustre clients accessing contended locks
4350 Details : if a client is granted a lock or receives a completion AST
4351 with a blocking AST already set it would not reply to the AST
4352 for LDLM_FL_CANCEL_ON_BLOCK locks. It now replies to such ASTs.
4355 Frequency : lfs setstripe, only systems with more than 160 OSTs
4357 Description: unable to set striping with a starting offset beyond OST 160
4358 Details : llapi_create_file() incorrectly limited the starting stripe
4359 index to the maximum single-file stripe count.
4362 Frequency : LDAP users only
4364 Description: lconf did not handle in-kernel recovery with LDAP properly
4365 Details : lconf/LustreDB get_refs() is searching the wrong namespace
4367 Severity : enhancement
4369 Description: bind OST threads to NUMA nodes to improve performance
4370 Details : all OST threads are uniformly bound to CPUs on a single NUMA
4371 node and do their allocations there to localize memory access
4373 Severity : enhancement
4375 Description: llmount can determine client NID directly from Myrinet (GM)
4376 Details : the client NID code from gmnalnid was moved directly into
4377 llmount, removing the need to use this or specifying the
4378 client NID explicitly when mounting GM clients with zeroconf
4381 Frequency : if client is started with down MDS
4383 Description: if client is started with down MDS mount hangs in ptlrpc_queue_wait
4384 Details : Having an LWI_INTR() wait event (interruptible, but no timeout)
4385 will wait indefinitely in ptlrpc_queue_wait->l_wait_event() after
4386 ptlrpc_import_delayed_req() because we didn't check if the
4387 request was interrupted, and we also didn't break out of the
4388 event loop if there was no timeout
4393 Description: data loss during non-page-aligned writes to a single file from
4394 both multiple nodes and multiple threads on one node at same time
4395 Details : updates to KMS and lsm weren't protected by common lock. Resulting
4396 inconsistency led to false short-reads, that were cached and later
4397 used by ->prepare_write() to fill in partially written page,
4398 leading to data loss.
4401 Frequency : always, if lconf --abort_recovery used
4403 Description: lconf --abort_recovery fails with 'Operation not supported'
4404 Details : lconf was attempting to abort recovery on the MDT device and not
4407 Severity : enhancement
4409 Description: remove cleanup logs
4410 Details : replace lconf-generated cleanup logs with lustre internal
4411 cleanup routines. Eliminates the need for client-cleanup and
4414 Severity : enhancement
4416 Description: add support for EAs (user and system) on lustre filesystems
4417 Details : it is now possible to store extended attributes in the Lustre
4418 client filesystem, and with the user_xattr mount option it
4419 is possible to allow users to store EAs on their files also
4421 Severity : enhancement
4423 Description: Add possibility (config option) to show minimal available OST free
4425 Details : When compiled with --enable-mindf configure option, statfs(2)
4426 (and so, df) will return least minimal free space available from
4427 all OSTs as amount of free space on FS, instead of summary of
4428 free spaces of all OSTs.
4430 Severity : enhancement
4432 Description: do not expand extent locks acquired on OST-side
4433 Details : Modify ldlm_extent_policy() to not expand local locks, acquired
4434 by server: they are not cached anyway.
4437 Frequency : when mmap is used/binaries executed from Lustre
4439 Description: Unmmap pages before throwing them away from read cache.
4440 Details : llap_shrink cache now attempts to unmap pages before discarding
4441 them (if unmapping failed - do not discard). SLES9 kernel has
4442 extra checks that trigger if this unmapping is not done first.
4447 Description: lconf didn't resolve symlinks before checking to see whether a
4448 given mountpoint was already in use
4451 Frequency : when migrating failover services
4452 Bugzilla : 6395, 9514
4453 Description: When migrating a subset of services from a node (e.g. failback
4454 from a failover service node) the remaining services would
4455 time out and evict clients.
4456 Details : lconf --force (implied by --failover) sets the global obd_timeout
4457 to 5 seconds in order to quickly disconnect, but this caused
4458 other RPCs to time out too quickly. Do not change the global
4459 obd_timeout for force cleanup, only set it for DISCONNECT RPCs.
4461 Severity : enhancement
4462 Frequency : if MDS is started with down OST
4463 Bugzilla : 9439,5706
4464 Description: Allow startup/shutdown of an MDS without depending on the
4465 availability of the OSTs.
4466 Details : Asynchronously call mds_lov_synchronize during MDS startup.
4467 Add appropriate locking and lov-osc refcounts for safe
4468 cleaning. Add osc abort_inflight calls in case the
4469 synchronize never started.
4472 Frequency : occasional (Cray XT3 only)
4474 Description: root not authorized to access files in CRAY_PORTALS environment
4475 Details : The client process capabilities were not honoured on the MDS in
4476 a CRAY_PORTALS/CRAY_XT3 environment. If the file had previously
4477 been accessed by an authorized user then root was able to access
4478 the file on the local client also. The root user capabilities
4479 are now allowed on the MDS, as this environment has secure UID.
4482 Frequency : occasional
4484 Description: ldiskfs "too long searching" message happens too often
4485 Details : A debugging message (otherwise harmless) prints too often on
4486 the OST console. This has been reduced to only happen when
4487 there are fragmentation problems on the filesystem.
4492 Description: Division by zero in statfs when all OSCs are inactive
4493 Details : lov_get_stripecnt() returns zero due to incorrect order of checks,
4494 lov_statfs divides by value returned by lov_get_stripecnt().
4498 Bugzilla : 9489, 3273
4499 Description: First write from each client to each OST was only 4kB in size,
4500 to initialize client writeback cache, which caused sub-optimal
4501 RPCs and poor layout on disk for the first writen file.
4502 Details : Clients now request an initial cache grant at (re)connect time
4503 and so that they can start streaming writes to the cache right
4504 away and always do full-sized RPCs if there is enough data.
4505 If the OST is rebooted the client also re-establishes its grant
4506 so that client cached writes will be honoured under the grant.
4511 Description: Slow ls (and stat(2) syscall) on files residing on IO-loaded OSTs
4512 Details : Now I/O RPCs go to different portal number and (presumably) fast
4513 lock requests (and glimses) and other RPCs get their own service
4514 threads pool that should be able to service those RPCs
4517 Severity : enhancement
4519 Description: Ability to exchange lustre version between client and servers and
4520 issue warnings at client side if client is too old. Also for
4521 liblustre clients there is ability to refuse connection of too old
4523 Details : New 'version' field is added to connect data structure that is
4524 filled with version info. That info is later checked by server and
4528 Frequency : rare, liblustre only.
4529 Bugzilla : 9296, 9581
4530 Description: Two simultaneous writes from liblustre at offset within same page
4531 might proceed at the same time overwriting eachother with stale
4533 Details : I/O lock withing llu_file_prwv was released too early, before data
4534 actually was hitting the wire. Extended lock-holding time until
4535 server acknowledges receiving data.
4538 Frequency : extremely rare. Never observed in practice.
4540 Description: avoid generating lustre_handle cookie of 0.
4541 Details : class_handle_hash() generates handle cookies by incrementing
4542 global counter, and can hit 0 occasionaly (this is unlikely, but
4543 not impossible, because initial value of cookie counter is
4544 selected randonly). Value of 0 is used as a sentinel meaning
4545 "unassigned handle" --- avoid it. Also coalesce two critical
4546 sections in this function into one.
4548 Severity : enhancement
4550 Description: allow liblustre clients to delegate truncate locking to OST
4551 Details : To avoid overhead of locking, liblustre client instructs OST to
4552 take extent lock in ost_punch() on client's behalf. New connection
4553 flag is added to handle backward compatibility.
4555 Severity : enhancement
4556 Bugzilla : 4928, 7341, 9758
4557 Description: allow number of OST service threads to be specified
4558 Details : a module parameter allows the number of OST service threads
4559 to be specified via "options ost ost_num_threads={N}" in the
4560 OSS's /etc/modules.conf or /etc/modprobe.conf.
4564 Bugzilla : 6146, 9635, 9895
4565 Description: servers crash with bad pointer in target_handle_connect()
4566 Details : In rare cases when a client is reconnecting it was possible that
4567 the connection request was the last reference for that export.
4568 We would temporarily drop the export reference and get a new
4569 one, but this may have been the last reference and the export
4570 was just destroyed. Get new reference before dropping old one.
4572 Severity : enhancement
4573 Frequency : if client is started with failover MDS
4575 Description: Allow multiple MDS hostnames in the mount command
4576 Details : Try to read the configuration from all specified MDS
4577 hostnames during a client mount in case the "primary"
4580 Severity : enhancement
4582 Description: Stop sending data to evicted clients as soon as possible.
4583 Details : Check if the client we are about to send or are sending data to
4584 was evicted already. (Check is done every second of waiting,
4585 for which l_wait_event interface was extended to allow checking
4586 of exit condition at specified intervals).
4589 Frequency : rare, normally only when NFS exporting is done from client
4591 Description: 'bad disk LOV MAGIC: 0x00000000' error when chown'ing files
4593 Details : Make mds_get_md() recognise empty md case and set lmm size to 0.
4596 Frequency : always, if srand() is called before liblustre initialization
4598 Description: Liblustre uses system PRNG disturbing its usage by user application
4599 Details : Introduce internal to lustre fast and high-quality PRNG for
4600 lustre usage and make liblustre and some other places in generic
4601 lustre code to use it.
4603 Severity : enhancement
4604 Bugzilla : 9477, 9557, 9870
4605 Description: Verify that the MDS configuration logs are updated when xml is
4606 Details : Check if the .xml configuration logs are newer than the config
4607 logs stored on the MDS and report an error if this is the case.
4608 Request --write-conf, or allow starting with --old_conf.
4610 Severity : enhancement
4612 Description: Handle symlinks in the path when checking if Lustre is mounted.
4613 Details : Resolve intermediate symlinks when checking if a client has
4614 mounted a filesystem to avoid duplicate client mounts.
4619 Description: lconf can hit an error exception but still return success.
4620 Details : The lconf command catches the Command error exception at the top
4621 level script context and will exit with the associated exit
4622 status, but doesn't ensure that this exit status is non-zero.
4627 Description: failure of ptlrpc thread startup can cause oops
4628 Details : Starting a ptlrpc service thread can fail if there are a large
4629 number of threads or the server memory is very fragmented.
4630 Handle this without oopsing.
4633 Frequency : always, only if liblustre and non-default acceptor port was used
4635 Description: liblustre cannot connect to servers with non-default acceptor port
4636 Details : tcpnal_set_default_params() was not called and was therefore
4637 ignoring the environment varaible TCPNAL_PORT, as well as other
4638 TCPNAL_ environment variables
4643 Description: two objects could be created on the same OST for a single file
4644 Details : If an OST is down, in some cases it was possible to create two
4645 objects on a single OST for a single file. No problems other
4646 than potential performance impact and spurious error messages.
4650 Bugzilla : 5681, 9562
4651 Description: Client may oops in ll_unhash_aliases
4652 Details : Client dcache may become inconsistent in race condition.
4653 In some cases "getcwd" can fail if the current directory is
4659 Description: Inode refcounting problems in NFS export code
4660 Details : link_raw functions used to call d_instantiate without obtaining
4661 extra inode reference first.
4665 Bugzilla : 9942, 9903
4666 Description: Referencing freed requests leading to crash, memleaks with NFS.
4667 Details : We used to require that call to ll_revalidate_it was always
4668 followed by ll_lookup_it. Also with revalidate_special() it is
4669 possible to call ll_revalidate_it() twice for the same dentry
4670 even if first occurence returned success. This fix changes semantic
4671 between DISP_ENQ_COMPLETE disposition flag to mean there is extra
4672 reference on a request referred from the intent.
4673 ll_intent_release() then releases such a request.
4676 Frequency : rare, normally benchmark loads only
4678 Description: unlinked inodes were kept in memory on the client
4679 Details : If a client is repeatedly creating and unlinking files it
4680 can accumulate a lot of stale inodes in the inode slab cache.
4681 If there is no other client load running this can cause the
4682 client node to run out of memory. Instead flush old inodes
4683 from client cache that have the same inode number as a new inode.
4686 Frequency : SLES9 2.6.5 kernel and long filenames only
4687 Bugzilla : 9969, 10379
4688 Description: utime reports stale NFS file handle
4689 Details : SLES9 uses out-of-dentry names in some cases, which confused
4690 the lustre dentry revalidation. Change it to always use the
4694 Frequency : rare, unless heavy write-truncate concurrency is continuous
4695 Bugzilla : 4180, 6984, 7171, 9963, 9331
4696 Description: OST becomes very slow and/or deadlocked during object unlink
4697 Details : filter_destroy() was holding onto the parent directory lock
4698 while truncating+unlinking objects. For very large objects this
4699 may block other threads for a long time and slow overall OST
4700 responsiveness. It may also be possible to get a lock ordering
4701 deadlock in this case, or run out of journal credits because of
4702 the combined truncate+unlink. Solution is to do object truncate
4703 first in one transaction without parent lock, and then do the
4704 final unlink in a new transaction with the parent lock. This
4705 reduces the lock hold time dramatically.
4708 Frequency : rare, 2.4 kernels only
4710 Description: MDS or OST cleanup may trip kernel BUG when dropping kernel lock
4711 Details : mds_cleanup() and filter_cleanup() need to drop the kernel lock
4712 before unmounting their filesystem in order to avoid deadlock.
4713 The kernel_locked() function in 2.4 kernels only checks whether
4714 the kernel lock is held, not whether it is this process that is
4715 holding it as 2.6 kernels do.
4720 Description: MDS or OST may oops/LBUG if a client is connecting multiple times
4721 Details : The client ptlrpc code may be trying to reconnect to a down
4722 server before a previous connection attempt has timed out.
4723 Increase the reconnect interval to be longer than the connection
4724 timeout interval to avoid sending duplicate connections to
4728 Frequency : echo_client brw_test command
4730 Description: fix echo_client to work with OST preallocated code
4731 Details : OST preallocation code (5137) didn't take echo_client IO path
4732 into account: echo_client calls filter methods outside of any
4733 OST thread and, hence, there is no per-thread preallocated
4734 pages and buffers to use. Solution: hijack pga pages for IO. As
4735 a byproduct, this avoids unnecessary data copying.
4739 Bugzilla : 3555, 5962, 6025, 6155, 6296, 9574
4740 Description: Client can oops in mdc_commit_close() after open replay
4741 Details : It was possible for the MDS to return an open request with no
4742 transaction number in mds_finish_transno() if the client was
4743 evicted, but without actually returning an error. Clients
4744 would later try to replay that open and may trip an assertion
4745 Simplify the client close codepath, and always return an error
4746 from the MDS in case the open is not successful.
4749 Frequency : rare, 2.6 OSTs only
4751 Description: OST may deadlock under high load on fragmented files
4752 Details : If there was a heavy load and highly-fragmented OST filesystems
4753 it was possible to have all the OST threads deadlock waiting on
4754 allocation of biovecs, because the biovecs were not released
4755 until the entire RPC IO was completed. Instead, release biovecs
4756 as soon as they are complete to ensure forward IO progress.
4758 Severity : enhancement
4760 Description: Support for specifying external journal device at mount
4761 Details : If an OST or MDS device is formatted with an external journal
4762 device, this device major/minor is stored in the ext3 superblock
4763 and may not be valid for failover. Allow detecting and
4764 specifying the external journal at mount time.
4769 Description: Mounting an MDS with pending unlinked files may cause oops
4770 Details : target_finish_recovery() calls mds_postrecov() which returned
4771 the number of orphans unlinked. mds_lov_connect->mds_postsetup()
4772 considers this an error and immediately begins cleaning up the
4773 lov, just after starting the mds_lov process
4775 Severity : enhancement
4777 Description: Implement 'lfs df' to report actual free space on per-OST basis
4778 Details : Add sub-command 'df' on 'lfs' to report the disk space usage of
4779 MDS/OSDs. Usage: lfs df [-i][-h]. Command Options: '-i' to report
4780 usage of objects; '-h' to report in human readable format.
4782 ------------------------------------------------------------------------------
4784 2005-08-26 Cluster File Systems, Inc. <info@clusterfs.com>
4791 Description: Mounting an ldiskfs file system with mballoc may crash OST node.
4792 Details : ldiskfs mballoc code may reference an uninitialized buddy struct
4793 at startup during orphan unlinking. Instead, skip buddy update
4794 before setup, as it will be regenerated after recovery is complete.
4799 Description: If an OST is inactive, its locks might reference stale inodes.
4800 Details : lov_change_cbdata() must iterate over all namespaces, even if
4801 they are inactive to clear inode references from the lock.
4803 Severity : enhancement
4804 Frequency : occasional, if non-standard max_dirty_mb used
4806 Description: Client will block write RPCs if not enough grant
4807 Details : If a client has max_dirty_mb smaller than max_rpcs_in_flight,
4808 then the client will block writes while waiting for another RPC
4809 to complete instead of consuming its dirty limit. With change
4810 we get improved performance when max_dirty_mb is small.
4812 Severity : enhancement
4813 Bugzilla : 3389, 6253
4814 Description: Add support for supplementary groups on the MDS.
4815 Details : The MDS has an upcall /proc/fs/lustre/mds/{mds}/group_upcall
4816 (set to /usr/sbin/l_getgroups if enabled) which will do MDS-side
4817 lookups for user supplementary groups into a cache.
4821 Description: O_CREAT|O_EXCL open flags in liblustre always return -EEXIST
4822 Details : Make libsysio to not enforce O_EXCL by clearing the flag,
4823 for liblustre O_EXCL is enforced by MDS.
4827 Description: readdir never returns NULL in liblustre.
4828 Details : Corrected llu_iop_getdirentries logic, to return offset of next
4829 dentry in struct dirent.
4833 Frequency : liblustre only, depends on application IO pattern
4834 Description: liblustre clients evicted if not contacting servers
4835 Details : Don't put liblustre clients into the ping_evictor list, so
4836 they will not be evicted by the pinger ever.
4838 Severity : enhancement
4840 Description: Add ability to evict clients by NID from MDS.
4841 Details : By echoing "nid:$NID" string into
4842 /proc/fs/lustre/mds/.../evict_client client with nid that equals to
4843 $NID would be instantly evicted from this MDS and from all active
4844 OSTs connected to it.
4848 Description: Do not query file size twice, somewhat slowing stat(2) calls.
4849 Details : lookup_it_finish() used to query file size from OSTs that was not
4854 Description: service threads change working directory to that of init
4855 Details : Starting lustre service threads may pin the working directory
4856 of the parent thread, making that filesystem busy. Threads
4857 now change to the working directory of init to avoid this.
4861 Frequency : during shutdown only
4862 Description: shutdown with a failed MDS or OST can cause unmount to hang
4863 Details : Don't resend DISCONNECT messages in ptlrpc_disconnect_import()
4868 Frequency : 2.6 only
4869 Description: chmod/chown may include an extra supplementary group
4870 Details : ll{,u}_mdc_pack_op_data() does not properly initialize the
4871 supplementary group and if none is specified this is used.
4874 Bugzilla : 5479 (6816)
4876 Description: Racing open + rm can assert client in mdc_set_open_replay_data()
4877 Details : If lookup is in progress on a file that is unlinked we might try
4878 to revalidate the inode and fail in revalidate after lookup is
4879 complete and ll_file_open() enqueues the open again but
4880 it_open_error() was not checking DISP_OPEN_OPEN errors correctly.
4883 Frequency : always, if lconf --abort_recovery used
4885 Description: lconf --abort_recovery fails with 'Operation not supported'
4886 Details : lconf was attempting to abort recovery on the MDT device and not
4889 ------------------------------------------------------------------------------
4891 2005-08-08 Cluster File Systems, Inc. <info@clusterfs.com>
4896 Frequency : rare (only unsupported configurations with a node running as an
4898 Bugzilla : 6514, 5137
4899 Description: Mounting a Lustre file system on a node running as an OST could
4901 Details : OSTs now preallocates memory needed to write out data at
4902 startup, instead of when needed, to avoid having to
4903 allocate memory in possibly low memory situations.
4904 Specifically, if the file system is mounted on on OST,
4905 memory pressure could force it to try to write out data,
4906 which it needed to allocate memory to do. Due to the low
4907 memory, it would be unable to do so and the node would
4908 become unresponsive.
4910 Severity : enhancement
4912 Description: Addition of lconf --service command line option
4913 Details : lconf now accepts a '--service <arg>' option, which is
4914 shorthand for 'lconf --group <arg> --select <arg>=<hostname>'
4916 Severity : enhancement
4918 Description: Failover mode is now the default for OSTs.
4919 Details : By default, OSTs will now run in failover mode. To return to
4920 the old behaviour, add '--failout' to the lmc line for OSTs.
4922 Severity : enhancement
4924 Description: Health checks are now provided for MDS and OSTs
4925 Details : Additional detailed health check information on MSD and OSTs
4926 is now provided through the procfs health_check value.
4929 Frequency : occasional, depends on IO load
4931 Description: Disk fragmentation on the OSTs could eventually cause slowdowns
4932 after numerous create/delete cycles
4933 Details : The ext3 inode allocation policy would not allocate new inodes
4934 very well on the OSTs because there are no new directories
4935 being created. Instead we look for groups with free space if
4936 the parent directories are nearly full.
4941 Description: Network or server problems during mount may cause partially
4942 mounted clients instead of returning an error.
4943 Details : The config llog parsing code may overwrite the error return
4944 code during mount error handling, returning success instead
4950 Description: MDS can fail to allocate large reply buffers
4951 Details : After long uptimes the MDS can fail to allocate large reply
4952 buffers (e.g. zconf client mount config records) due to memory
4953 fragmentation or consumption by the buffer cache. Preallocate
4954 some large reply buffers so that these replies can be sent even
4955 under memory pressure.
4959 Frequency : rare (liblustre)
4960 Description: fsx running with liblustre complained that using truncate() to
4961 extend the file doesn't work. This patch corrects that issue.
4962 Details : This is the liblustre equivalent of the fix for bug 6196. Fixes
4963 ATTR_SIZE and lsm use in llu_setattr_raw.
4967 Frequency : rare, only 2.6 kernels
4968 Description: Unusual file access patterns on the MDS may result in inode
4969 data being lost in very rare circumstances.
4970 Details : Bad interaction between the ea-in-inode patch and the "no-read"
4971 code in the 2.6 kernel caused the inode and/or EA data not to
4972 be read from disk, causing single-file corruption.
4976 Frequency : rare, only 2.6 filesystems using extents
4977 Description: Heavy concurrent write and delete load may cause data corruption.
4978 Details : It was possible under high-load situations to have an extent
4979 metadata block in the block device cache from a just-unlinked
4980 file overwrite a newly-allocated data block. We now unmap any
4981 metadata buffers that alias just-allocated data blocks.
4985 Frequency : filesystems with default stripe_count larger than 77
4986 Description: lconf+mke2fs fail when formatting filesystem with > 77 stripes
4987 Details : lconf specifies an inode size of 4096 bytes when the default
4988 stripe_count is larger than 77. This conflicts with the default
4989 inode density of 1 per 4096 bytes. Allocate smaller inodes in
4990 this case to avoid pinning too much memory for large EAs.
4992 ------------------------------------------------------------------------------
4994 2005-07-07 Cluster File Systems, Inc. <info@clusterfs.com>
4999 Frequency : rare (extremely heavy IO load with hundreds of clients)
5001 Description: Client is evicted, gets IO error writing to file
5002 Details : lock ordering changes for bug 5492 reintroduced bug 3267 and
5003 caused clients to be evicted for AST timeouts. The fixes in
5004 bug 5192 mean we no longer need to have such short AST timeouts
5005 so ldlm_timeout has been increased.
5008 Frequency : occasional during --force or --failover shutdown under load
5009 Bugzilla : 5949, 4834
5010 Description: Server oops/LBUG if stopped with --force or --failover under load
5011 Details : a collection of import/export refcount and cleanup ordering
5012 issues fixed for safer force cleanup
5015 Frequency : only filesystems larger than 120 OSTs
5016 Bugzilla : 5990, 6223
5017 Description: lfs getstripe would oops on a very large filesystem
5018 Details : lov_getconfig used kfree on vmalloc'd memory
5021 Frequency : only filesystems exporting via NFS to Solaris 10 clients
5022 Bugzilla : 6242, 6243
5023 Description: reading from files that had been truncated to a non-zero size
5024 but never opened returned no data
5025 Details : ll_file_read() reads zeros from no-object files to EOF
5030 Description: A bug in MDS/OSS recovery could cause the OSS to fail an assertion
5031 Details : There's little harm in aborting MDS/OSS recovery and letting it
5032 try again, so I removed the LASSERT and return an error instead.
5034 Severity : enhancement
5036 Description: New debugging infrastructure for tracking down data corruption
5037 Details : The I/O checksum code was replaced to: (a) control it at runtime,
5038 (b) cover more of the client-side code path, and (c) try to narrow
5039 down where problems occurred
5043 Bugzilla : 3819, 4364, 4397, 6313
5044 Description: Racing close and eviction MDS could cause assertion in mds_close
5045 Details : It was possible to get multiple mfd references during close and
5046 client eviction, leading to one thread referencing a freed mfd.
5048 Severity: : enhancement
5049 Bugzilla : 3262, 6359
5050 Description: Attempts to reconnect to servers are now more aggressive.
5051 Details : This builds on the enhanced upcall-less recovery that was added
5052 in 1.4.2. When trying to reconnect to servers, clients will
5053 now try each server in the failover group every 10 seconds. By
5054 default, clients would previously try one server every 25 seconds.
5059 Description: After recovery, certain operations trigger a failed
5060 assertion on a client.
5061 Details : Failing over an mds, using lconf -d --failover, while a
5062 client was doing a readdir() call would cause the client to
5063 LBUG after recovery completed and the readdir() was resent.
5065 Severity : enhancement
5067 Description: Default groups are now added by lconf
5068 Details : You can now run lconf --group <servicename> without having to
5069 manually add groups with lmc.
5072 Frequency : occasional
5074 Description: Nodes with an elan id of 0 trigger a failed assertion
5077 Frequency : always when accessing e.g. tty/console device nodes
5079 Description: tty and some other devices nodes cannot be used on lustre
5080 Details : file's private_data field is used by device data and lustre
5081 values in there got lost. New field was added to struct file to
5082 store fs-specific private data.
5085 Frequency : when exporting Lustre via NFS
5087 Description: NFSD failed occasionally when looking up a path component
5088 Details : NFSD is looking up ".." which was broken in ext3 directories
5089 that had grown large enough to become hashed.
5092 Frequency : Clusters with multiple interfaces not on the same subnet
5094 Description: Nodes will repeatedly try to reconnect to an interface which it
5095 cannot reach and report an error to the log.
5096 Details : Extra peer list entries will be created by lconf with some peers
5097 unreachable. lconf now validates the peer before adding it.
5100 Frequency : Only if a default stripe is set on the filesystem root.
5102 Description: Setting a default stripe on the filesystem root prevented the
5103 filesystem from being remounted.
5104 Details : The client was sending extra request flags in the root getattr
5105 request and did not allocate a reply buffer for the dir EA.
5108 Frequency : occasional, higher if lots of files are accessed by one client
5109 Bugzilla : 6159, 6097
5110 Description: Client trips assertion regarding lsm mismatch/magic
5111 Details : While revalidating inodes the VFS looks up inodes with ifind()
5112 and in rare cases can find an inode that is being freed.
5113 The ll_test_inode() code will free the lsm during ifind()
5114 when it finds an existing inode and then the VFS later attaches
5115 this free lsm to a new inode.
5119 Bugzilla : 6422, 7030
5120 Description: MDS deadlock between mkdir and client eviction
5121 Details : Creating a new file via mkdir or mknod (starting a transaction
5122 and getting the ns lock) can deadlock with client eviction
5123 (gets ns lock and trying to finish a synchronous transaction).
5126 Frequency : occasional
5127 Description: While starting a server, the fsfilt_ext3 module could not be
5129 Details : CFS's improved ext3 filesystem is named ldiskfs for 2.6
5130 kernels. Previously, lconf would still use the ext3 name
5131 when trying to load modules. Now, it will correctly use
5132 ext3 on 2.4 and ldiskfs on 2.6.
5134 Severity : enhancement
5135 Description: The default stripe count has been changed to 1
5136 Details : The interpretation of the default stripe count (0, to lfs
5137 or lmc) has been changed to mean striping across a single
5138 OST, rather than all available. For general usage we have
5139 found a stripe count of 1 or 2 works best.
5141 Severity : enhancement
5142 Description: Add support for compiling against Cray portals.
5143 Details : Conditional compiling for some areas that are different
5147 Frequency : occasional
5148 Bugzilla : 6409, 6834
5149 Description: Creating files with an explicit stripe count may lead to
5150 a failed assertion on the MDS
5151 Details : If some OSTs are full or unavailable, creating files may
5152 trigger a failed assertion on the MDS. Now, Lustre will
5153 try to use other servers or return an error to the
5157 Frequency : occasional
5159 Description: Multiple concurrent overlapping read+write on multiple SMP nodes
5160 caused lock timeout during readahead (since 1.4.2).
5161 Details : Processes doing readahead might match a lock that hasn't been
5162 granted yet if there are overlapping and conflicting lock
5163 requests. The readahead process waits on ungranted lock
5164 (original lock is CBPENDING), while OST waits for that process
5165 to cancel CBPENDING read lock and eventually evicts client.
5167 Severity : enhancement
5169 Description: Initial enabling of flock support for clients
5170 Details : Implements fcntl advisory locking and file status functions.
5171 This feature is provided as an optional mount flag (default
5172 off), and is NOT CURRENTLY SUPPORTED. Not all types of record
5173 locking are implemented yet, and those that are are not guaranteed
5174 to be completely correct in production environments.
5175 mount -t lustre -o [flock|noflock] ...
5178 Frequency : occasional
5180 Description: OSTs running 2.4 kernels but with extents enabled might trip an
5181 assertion in the ext3 JBD (journaling) layer.
5182 Details : The b_committed_data struct is protected by the big kernel lock
5183 in 2.4 kernels, serializing journal_commit_transaction() and
5184 ext3_get_block_handle->ext3_new_block->find_next_usable_block()
5185 access to this struct. In 2.6 kernels there is finer grained
5186 locking to improve SMP performance of the JBD layer.
5190 Description: Changes the "SCSI I/O Stats" kernel patch to default to "enabled"
5192 -----------------------------------------------------------------------------
5194 2005-05-05 Cluster File Systems, Inc. <info@clusterfs.com>
5196 NOTE: Lustre 1.4.2 uses an incompatible network protocol than previous
5197 versions of Lustre. Please update all servers and clients to
5198 version 1.4.2 or later at the same time. You must also run
5199 "lconf --write-conf {config}.xml" on the MDS while it is stopped
5200 to update the configuration logs.
5202 - fix for HPUX NFS client breakage when NFS exporting Lustre (5781)
5203 - mdc_enqueue does not need max_mds_easize request buffer on send (5707)
5204 - swab llog records of type '0' so we get proper header size/idx (5861)
5205 - send llog cancel req to DLM cancel portal instead of cb portal (5515)
5206 - fix rename of one directory over another leaking an inode (5953)
5207 - avoid SetPageDirty on 2.6 (5981)
5208 - don't re-add just-being-destroyed locks to the waiting list (5653)
5209 - when creating new directories, inherit the parent's custom
5210 striping settings if present parent (3048)
5211 - flush buffers from cache before direct IO in 2.6 obdfilter (4982)
5212 - don't hold i_size_sem in ll_nopage() and ll_ap_refresh_count (6077)
5213 - don't hold client locks on temporary worklist from l_lru (5666)
5214 - handle IO errors in 2.6 obdfilter bio completion routine (6046)
5215 - automatically evict dead clients (5921)
5216 - Update file size properly in create+truncate+fstat case (6196)
5217 - Do not unhash mountpoint dentries, do not allow removal of
5219 - Avoid lock ordering deadlock issue with write/truncate (6203,5654)
5220 - reserve enough journal credits in fsfilt_start_log for setattr (4554)
5221 - ldlm_enqueue freed-export error path would always LBUG (6149,6184)
5222 - don't reference lr_lvb_data until after we hold lr_lvb_sem (6170)
5223 - don't overwrite last_rcvd if there is a *_client_add() error (6086)
5224 - Correctly handle reads of files with no objects (6243)
5225 - lctl recover will also mark a device active if deactivate used (5933)
5227 - by default create 1 inode per 4kB space on MDS, per 16kB on OSTs
5228 - allow --write-conf on an MDS with different nettype than client (5619)
5229 - don't write config llogs to MDS for mounts not from that MDS (5617)
5230 - lconf should create multiple TCP connections from a client (5201)
5231 - init scripts are now turned off by default; run chkconfig --on
5232 lustre and chkconfig --on lustrefs to use them
5233 - upcalls are no longer needed for clients to recover to failover
5235 - add --abort-recovery option to lconf to abort recovery on device
5237 - add support for an arbitrary number of OSTs (3026)
5238 - Quota support protocol changes.
5239 - forward compatibility changes to wire structs (6007)
5240 - rmmod NALs that might be loaded because of /etc/modules.conf (6133)
5241 - support for mountfsoptions and clientoptions to the Lustre LDAP (5873)
5242 - improved "lustre status" script
5243 - initialize blocksize for non-regular files (6062)
5244 - added --disable-server and --disable-client configure options (5782)
5245 - introduce a lookup cache for lconf to avoid repeated DB scans (6204)
5246 - Vanilla 2.4.29 support
5247 - increase maximum number of obd devices to 520 (6242)
5248 - remove the tcp-zero-copy patch from the suse-2.4 series (5902)
5249 - Quadrics Elan drivers are now included for the RHEL 3 2.4.21 and
5250 SLES 9 2.6.5 kernels
5251 - limit stripes per file to 160 (the maximum EA size) (6093)
5253 2005-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5256 - don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
5257 - hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
5258 - fix setattr mtime regression from lovcleanup merge (4829, 5669)
5259 - workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
5260 - small ext3 extents cleanups and fixes (5733)
5261 - improved mballoc code, several small races and bugs fixed (5733, 5638)
5262 - kernel version 43 - fix remove_suid bugs in both 2.4 and 2.6 (5695)
5263 - avoid needless client->OST connect, fix handle mismatch (5317)
5264 - fix DLM error path that led to out-of-sync client, long delays (5779)
5265 - support common vfs-enforced mount options (nodev,nosuid,noexec) (5637)
5266 - fix several locking issues related to i_size (5492,5624,5654,5672)
5267 - don't move pending lock onto export if it is already evicted (5683)
5268 - fix kernel oops when creating .foo in unlinked directory (5548)
5269 - fix deadlock in obdfilter statistics vs. object create (5811)
5270 - use time_{before,after} to avoid timer jiffies wrap (5882)
5271 - shutdown --force/--failover stability (3607,3651,4797,5203,4834)
5272 - Do not leak request if server was not able to process it (5154)
5273 - If mds_open unable to find parent dir, make that negative lookup(5154)
5274 - don't create new directories with extent-mapping (5909, 5936)
5276 - fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
5277 - don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
5278 - change obd_increase_kms to obd_adjust_kms (up or down) (5654)
5279 - lconf, lmc search both /usr/lib and /usr/lib64 for Python libs (5800)
5280 - support for RHEL4 kernel on i686 (5773)
5281 - provide error messages when incompatible logs are encountered (5898)
5283 2005-02-18 Cluster File Systems, Inc. <info@clusterfs.com>
5284 * version 1.4.0.10 (1.4.1 release candidate 1)
5286 - don't keep a lock reference when lock is not granted (4238)
5287 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5288 - add per-fs limit of Lustre pages in page cache, avoid OOM (4699)
5289 - drop import inflight refcount on signal_completed_replay error (5255)
5290 - unlock page after async write error during send (3677)
5291 - handle missing objects in filter_preprw_read properly (5265)
5292 - no transno return for symlink open, don't save no-trasno open (3440)
5293 - don't try to complete elan receive that already failed (4012)
5294 - free RPC server reply state on error (5406)
5295 - clean up thread from ptlrpc_start_thread() on error (5160)
5296 - readahead could read extra page into cache that wasn't ejected (5388)
5297 - prevent races in class_attach/setup/cleanup/detach (5260)
5298 - don't dereference de->d_inode after l_dput of de (5458)
5299 - use "int" for stripe value returned from lock_to_stripe (5544)
5300 - mballoc allocation and error-checking fixes in 2.6 (5504)
5301 - block device patches to fix I/O request sizes in 2.6 (5482)
5302 - look up hostnames for IB nals (5602)
5303 - 2.6 changed lock ordering of 2 semaphores, caused deadlock (5654)
5304 - don't start multiple acceptors for the same port (5277)
5305 - fix incorrect LASSERT in mds_getattr_name (5635)
5306 - export a proc file for general "ping" checking (5628)
5307 - fix "lfs check" to not block when the MDS is down (5628)
5309 - service request history (4965)
5310 - put {ll,lov,osc}_async_page structs in a single slab (4699)
5311 - create an "evict_client" /proc entry on OSTs, like the MDS has
5312 - fix mount usage message, return errors per mount(8) (5168)
5313 - change grep [] to grep "[]" in tests so they work in more UMLs
5314 - fix ppc64/x86_64 spec to use %{_libdir} instead of /usr/lib (5389)
5315 - remove ancient LOV_MAGIC_V0 EA support (5047)
5316 - add "disk I/Os in flight" and "I/O req time" stats in obdfilter
5317 - align r/w RPCs to PTLRPC_MAX_BRW_SIZE boundary for performance (3451)
5318 - allow readahead allocations to fail when low on memory (5383)
5319 - mmap locking landed again, after considerable improvement (2828)
5320 - add get_hostaddr() to lustreDB.py for LDAP support (5459)
5322 2004-11-23 Cluster File Systems, Inc. <info@clusterfs.com>
5325 - send OST transaction number in read/write reply to free req (4966)
5326 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5327 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5328 - the watchdog thread now runs as interruptible (5246)
5329 - flock/lockf fixes (but it's still disabled, pending 5135)
5330 - don't use EXT3 constants in llite code (5094)
5331 - memory shortage at startup could cause assertion (5176)
5333 - reorganization of lov code
5334 - single portals codebase
5336 - add extents/mballoc support (5025)
5337 - direct I/O reads in the obdfilter (4048)
5338 - kernel patches from LNXI for 2.6 (bluesmoke, perfctr, mtd, kexec)
5340 tbd Cluster File Systems, Inc. <info@clusterfs.com>
5343 - send OST transaction number in read/write reply to free req (4966)
5344 - don't ASSERT in ptl_send_rpc() if we run out of memory (5119)
5345 - lock /proc/sys/portals/routes internal state, avoiding oops (4827)
5346 - the watchdog thread now runs as interruptible (5246)
5347 - handle missing objects in filter_preprw_read properly (5265)
5348 - unsafe list practices (rarely) led to infinite eviction loop (4908)
5349 - drop import inflight refcount on signal_completed_replay error (5255)
5350 - unlock page after async write error during send (3677)
5351 - return original error code on reconstructed replies (3761)
5352 - no transno return for symlink open, don't save no-trasno open (3440)
5354 - add pid to ldlm debugging output (4922)
5355 - bump the watchdog timeouts -- we can't handle 30sec yet
5356 - extra debugging for orphan dentry/inode bug (5259)
5358 2004-11-16 Cluster File Systems, Inc. <info@clusterfs.com>
5361 - fix TCP_NODELAY bug, which caused extreme perf regression (5134)
5362 - allocate qswnal tx descriptors singly to avoid fragmentation (4504)
5363 - don't LBUG on obdo_alloc() failure, use OBD_SLAB_ALLOC() (4800)
5364 - fix NULL dereference in /proc/sys/portals/routes (4827)
5365 - allow failed mdc_close() operations to be interrupted (4561)
5366 - stop precreate on OST before MDS would time out on it (4778)
5367 - don't send partial-page writes before EOF from client (4410)
5368 - discard client grant for sub-page writes on large-page clients (4520)
5369 - don't free dentries not owned by NFS code, check generation (4806)
5370 - fix lsm leak if mds_create_objects() fails (4801)
5371 - limit debug_daemon file size, always print CERROR messages (4789)
5372 - use transno after validating reply (3892)
5373 - process timed out requests if import state changes (3754)
5374 - update mtime on OST during writes, return in glimpse (4829)
5375 - add mkfsoptions to LDAP (4679)
5376 - use ->max_readahead method instead of zapping global ra (5039)
5377 - don't interrupt __l_wait_event() during strace
5379 - add software watchdogs to catch hung threads quickly (4941)
5380 - make lustrefs init script start after nfs is mounted
5381 - fix CWARN/ERROR duplication (4930)
5382 - return async write errors to application if possible (2248)
5383 - add /proc/sys/portal/memused (bytes allocated by PORTALS_ALLOC)
5384 - print NAL number in %x format (4645)
5385 - update barely-supported suse-2.4.21-171 series (4842)
5386 - support for sles 9 %post scripts
5387 - support for building 2.6 kernel-source packages
5388 - support for sles km_* packages
5390 2004-10-07 Cluster File Systems, Inc. <info@clusterfs.com>
5393 - ignore -ENOENT errors in osc_destroy (3639)
5394 - notify osc create thread that OSC is being cleaned up (4600)
5395 - add nettype argument for llmount in #5d in conf-sanity.sh (3936)
5396 - reconstruct ost_handle() like mds_handle() (4657)
5397 - create a new thread to do import eviction to avoid deadlock (3969)
5398 - let lconf resolve symlinked-to devices (4629)
5399 - don't unlink "objects" from directory with default EA (4554)
5400 - hold socknal file ref over connect in case target is down (4394)
5401 - allow more than 32000 subdirectories in a single directory (3244)
5402 - fix blocks count for O_DIRECT writes (3751)
5403 - OST returns ENOSPC from object create when no space left (4539)
5404 - don't send truncate RPC if file size isn't changing (4410)
5405 - limit OSC precreate to 1/2 of value OST considers bogus (4778)
5406 - bind to privileged port in socknal and tcpnal (3689)
5408 - rate limit CERROR/CWARN console message to avoid overload (4519)
5409 - GETFILEINFO dir ioctl returns LOV EA + MDS stat in 1 call (3327)
5410 - basic mmap support (3918)
5411 - kernel patch series update from b1_4 (4711)
5413 2004-09-16 Cluster File Systems, Inc. <info@clusterfs.com>
5416 - avoid crash during MDS cleanup with OST shut down (2775)
5417 - fix loi_list_lock/oig_lock inversion on interrupted IO (4136)
5418 - don't use bad inodes on the MDS (3744)
5419 - dynamic object preallocation to improve recovery speed (4236)
5420 - don't hold spinlock over lock dumping or change debug flags (4401)
5421 - don't zero obd_dev when it is force cleaned (3651)
5422 - print grants to console if they go negative (4431)
5423 - "lctl deactivate" will stop automatic recovery attempts (3406)
5424 - look for existing locks in ldlm_handle_enqueue() (3764)
5425 - don't resolve lock handle twice in recovery avoiding race (4401)
5426 - revalidate should check working dir is a directory (4134)
5428 - don't always mark "slow" obdfilter messages as errors (4418)
5430 2004-08-24 Cluster File Systems, Inc. <info@clusterfs.com>
5433 - don't close LustreDB during write_conf until it is done (3860)
5434 - fix typo in lconf for_each_profile (3821)
5435 - allow dumping logs from multiple threads at one time (3820)
5436 - don't allow multiple threads in OSC recovery (3812)
5437 - fix debug_size parameters (3864)
5438 - fix mds_postrecov to initialize import for llog ctxt (3121)
5439 - replace config semaphore with spinlock (3306)
5440 - be sure to send a reply for a CANCEL rpc with bad export (3863)
5441 - don't allow enqueue to complete on a destroyed export (3822)
5442 - down write_lock before checking llog header bitmap (3825)
5443 - recover from lock replay timeout (3764)
5444 - up llog sem before sending rpc (3652)
5445 - reduce ns lock hold times when setting kms (3267)
5446 - change a dlm LBUG to LASSERTF, to maybe learn something (4228)
5447 - fix NULL deref and obd_dev leak on setup error (3312)
5448 - replace some LBUG about llog ops with error handling (3841)
5449 - don't match INVALID dentries from d_lookup and spin (3784)
5450 - hold dcache_lock while marking dentries INVALID and hashing (4255)
5451 - fix invalid assertion in ptlrpc_set_wait (3880)
5453 - add libwrap support for the TCP acceptor (3996)
5454 - add /proc/sys/portals/routes for non-root route listing (3994)
5455 - allow setting MDS UUID in .xml (2580)
5456 - print the stack of a process that LBUGs (4228)
5458 2004-07-14 Cluster File Systems, Inc. <info@clusterfs.com>
5461 - don't cleanup request in ll_file_open() on failed MDS open (3430)
5462 - make sure to unset replay flag from failed open requests (3440)
5463 - if default stripe count is 0, use OST count for inode size (3636)
5464 - update parent mtime/ctime on client for create/unlink (2611)
5465 - drop dentry ref in ext3_add_link from open_connect_dentry (3266)
5466 - free recovery state on server during a forced cleanup (3571)
5467 - unregister_reply for resent reqs (3063)
5468 - loop back devices mounting and status check on 2.6 (3563)
5469 - fix resource-creation race that can provoke i_size == 0 (3513)
5470 - don't try to use bad inodes returned from MDS/OST fs lookup (3688)
5471 - more debugging for page-accounting assertion (3746)
5472 - return -ENOENT instead of asserting if ost getattr+unlink race (3558)
5473 - avoid deadlock after precreation failure (3758)
5474 - fix race and lock order deadlock in orphan handling (3450, 3750)
5475 - add validity checks when grabbing inodes from l_ast_data (3599)
5477 - add /proc/.../recovery_status to obdfilter (3428)
5478 - lightweight CDEBUG infrastructure, debug daemon (3668)
5479 - change default OSC RPC parameters to be better on small clusters
5480 - turn off OST read cache for files smaller than 32MB
5481 - install man pages and include them in rpms (3100)
5482 - add new init script for (un)mounting lustre filesystems (2593)
5483 - run chkconfig in %post for init scripts (3701)
5484 - drop scimac NAL (unmaintained)
5486 2004-06-17 Cluster File Systems, Inc. <info@clusterfs.com>
5489 - clean kiobufs before and after use (3485)
5490 - strip trailing '/'s before comparing paths with /proc/mounts (3486)
5491 - remove assertions to work around "in-flight rpcs" recovery bug (3063)
5492 - change init script to fail more clearly if not run as root (1528)
5493 - allow clients to reconnect during replay (1742)
5494 - fix ns_lock/i_sem lock ordering deadlock for kms update (3477)
5495 - don't do DNS lookups on NIDs too small for IP addresses (3442)
5496 - re-awaken ptlrpcd if new requests arrive during check_set (3554)
5497 - fix cond_resched (3554)
5498 - only evict unfinished clients after recovery (3515)
5499 - allow bulk resend, prevent data loss (3570)
5500 - dynamic ptlrpc request buffer allocation (2102)
5501 - don't allow unlinking open directory if it isn't empty (2904)
5502 - set MDS/OST threads to umask 0 to not clobber client modes (3359)
5503 - remove extraneous obd dereference causing LASSERT failure (3334)
5504 - don't use get_cycles() when creating temp. files on the mds (3156)
5505 - hold i_sem when setting i_size in ll_extent_lock() (3564)
5506 - handle EEXIST for set-stripe, set proper directory name (3336)
5508 - servers can dump a log evicting a client - lustre.dump_on_timeout=1
5509 - fix ksocknal_fmb_callback() error messages (2918)
5511 2004-05-27 Cluster File Systems, Inc. <info@clusterfs.com>
5514 - don't copy lvb into (possibly NULL) reply on error (2983)
5515 - don't deref dentry after dput, don't free lvb on error (2922)
5516 - use the kms to determine writeback rpc length (2947)
5517 - increment oti_logcookies when osc is inactive (2948)
5518 - update client's i_blocks count via lvb messages (2543)
5519 - handle intent open/close of special files properly (1557)
5520 - mount MDS with errors=remount-ro, like obdfilter (2009)
5521 - initialize lock handle to avoid ASSERT on error cleanup (3057)
5522 - don't use cancelling-locks' kms values (2947)
5523 - use highest lock extent for kms, not last one (2925)
5524 - don't dereference ERR_PTR() dentry in error handling path (3107)
5525 - fix thread race in portals_debug_dumplog() (3122)
5526 - create lprocfs device entries at setup instead of at attach (1519)
5527 - common AST error handler, don't evict client on completion race (3145)
5528 - zero nameidata in detach_mnt in 2.6 (3118)
5529 - verify d_inode after revalidate_special is valid in 2.6 (3116)
5530 - use lustre_put_super() to handle zconf unmounts in 2.6 (3064)
5531 - initialize RPC timeout timer earlier for 2.6 (3219)
5532 - don't dereference NULL reply buffer if mdc_close was never sent (2410)
5533 - print nal/nid for unknown nid (3258)
5534 - additional checks for oscc recovery before doing precreate (3284)
5535 - fix ll_extent_lock() error return code for 64-bit systems (3043)
5536 - don't crash in mdc_close for bad permissions on open (3285)
5537 - zero i_rdev for non-device files (3147)
5538 - clear page->private before handing to FS, better assertion (3119)
5539 - tune the read pipeline (3236)
5540 - fix incorrect decref of invalidated dentry (2350)
5541 - provide read-ahead stats and refine rpc in flight stats (3328)
5542 - don't hold journal transaction open across create RPC (3313)
5543 - update atime on MDS at close time (3265)
5544 - close LDAP connection when recovering to avoid server load (3315)
5545 - update iopen-2.6 patch with fixes from 2399,2517,2904 (3301)
5546 - don't leak open file on MDS after open resend (3325)
5547 - serialize filter_precreate and filter_destroy_precreated (3329)
5548 - loop device shouldn't call sync_dev() for nul device (3092)
5549 - clear page cache after eviction (2766)
5550 - resynchronize MDS->OST in background (2824)
5551 - refuse to mount the same filesystem twice on same mountpoint (3394)
5552 - allow llmount to create routes for mounting behind routers (3320)
5553 - push lock cancellation to blocking thread for glimpse ASTs (3409)
5554 - don't call osc_set_data_with_check() for TEST_LOCK matches (3159)
5555 - fix rare problem with rename on htree directories (3417)
5557 - allow default OST striping configuration per directory (1414)
5558 - fix compilation for qswnal for 2.6 kernels (3125)
5559 - increase maximum number of MDS request buffers for large systems
5560 - change liblustreapi to be useful for external progs like lfsck (3098)
5561 - increase local configuration timeout for slow disks (3353)
5562 - allow configuring ldlm AST timeout - lustre.ldlm_timeout=<seconds>
5564 2004-03-22 Cluster File Systems, Inc. <info@clusterfs.com>
5567 - fixes for glimpse AST timeouts / incorrectly 0-sized files (2818)
5568 - don't overwrite extent policy data in reply if lock was blocked (2901)
5569 - drop filter export grants atomically with removal from device (2663)
5570 - del obd_self_export from work_list in class_disconnect_exports (2908)
5571 - don't LBUG if MDS recovery times out during orphan cleanup (2530)
5572 - swab reply message in mdc_close, other PPC fixes (2464)
5573 - fix destroying of named logs (2325)
5574 - overwrite old logs when running lconf --write_conf (2264)
5575 - bump LLOG_CHUNKSIZE to 8k to allow for larger clusters (2306)
5576 - fix race in target_handle_connect (2898)
5577 - mds_reint_create() should take same inode create lock (2926)
5578 - correct journal credits calculated for CANCEL_UNLINK_LOG (2931)
5579 - don't close files for self_export to avoid uninitialized obd (2936)
5580 - allow MDS with the same name as client node (2939)
5581 - hold dentry reference for closed log files for unlink (2325)
5582 - reserve space for all logs during transactions (2059)
5583 - don't evict page beyond end of stripe extent (2925)
5584 - don't oops on a deleted current working directory (2399)
5585 - handle hard links to targets without a parent properly (2517)
5586 - don't dereference NULL lock when racing during eviction (2867)
5587 - don't grow lock extents when lots of conflicting locks (2919)
5589 2004-03-04 Cluster File Systems, Inc. <info@clusterfs.com>
5592 - account for cache space usage on clients to avoid data loss (974)
5593 - lfsck support in lustre kernel code (2349)
5594 - reduce journal credits needed for BRW writes (2370)
5595 - orphan handling to avoid losing space on client/server crashes
5596 - ptlrpcd can be blocked, stopping ALL progress (2477)
5597 - use lock value blocks to assist in proper KMS, faster stat (1021)
5598 - takes i_sem instead of DLM locks internally on obdfilter (2720)
5599 - recovery for initial connections (2355)
5600 - fixes for mds_cleanup_orphans (1934)
5601 - abort_recovery crashes MDS in b_eq (mds_unlink_orphan) (2584)
5602 - block all file creations until orphan recovery completes (1901)
5603 - client remove rq_connection from request struct (2423)
5604 - conf-sanity test_5, proper cleanup in umount log not availale (2640)
5605 - recovery timer race (2670)
5606 - mdc_close recovey bug (2532)
5607 - ptlrpc cleanup bug (2710)
5608 - mds timeout on local locks (2588)
5609 - namespace lock held during RPCs (2431)
5610 - handle interrupted sync write properly (2503)
5611 - don't try to handle a message that hasn't been replied to (2699)
5612 - client assert failure during cleanup after abort recovery (2701)
5613 - leak mdc device after failed mount (2712)
5614 - ptlrpc_check_set allows timedout requests to complete (2714)
5615 - wait for inflight reqs when ptlrpcd finishes (2710)
5616 - make sure unregistered services are removed from the srv_list
5617 - reset bulk XID's when resending them (caught by 1138 test)
5618 - unregister_bulk after timeout
5619 - fix lconf error (2694)
5620 - handle write after unfinished setstripe, stripe-only getstripe (2388)
5621 - readahead locks pages, leaves pending causing memory pressure (2673)
5622 - increase OST request buffers to 4096 on large machines (2729)
5623 - fix up permission of existing directories in simple_mkdir (2661)
5624 - init deleted item, add assertions ptlrpc_abort_inflight() (2725)
5625 - don't assign transno to errored transactions (2742)
5626 - don't delete objects on OST if given a bogus objid from MDS (2751)
5627 - handle large client PAGE_SIZE readdir on small PAGE_SIZE MDS (2777)
5628 - if rq_no_resend, then timeout request after recovery (2432)
5629 - fix MDS llog_logid record size, 64-bit array alignment (2733)
5630 - don't call usermode_helper from ptlrpcd, DEFAULT upcall (2773)
5631 - put magic in mount.lustre data, check for bad/NULL mount data (2529)
5632 - MDS recovery shouldn't delete objects that it has given out (2730)
5633 - if enqueue arrives after completion, don't clobber LVB (2819)
5634 - don't unlock pages twice when trigger_group_io returns error (2814)
5635 - don't deref NULL rq_repmsg if ldlm_handle_enqueue failed (2822)
5636 - don't write pages to disk if there was an error (1450)
5637 - don't ping imports that have recovery disabled (2676)
5638 - take buffered bytes into account when balancing socknal conn (2817)
5639 - hold a DLM lock over readdir always, use truncate_inode_pages (2706)
5640 - reconnect unlink llog connection after MDS reconnects to OST (2816)
5641 - remove little-endian swabbing of llog records (1987)
5642 - set/limit i_blksize to LL_MAX_BLKSIZE on client (2884)
5643 - retry reposting request buffers if they fail (1191)
5644 - grow extent at grant time to avoid granting a revoked lock (2809)
5645 - lock revoke doesn't evict page if covered by a second lock (2765)
5646 - disable VM readahead to avoid reading outside lock extents (2805)
5648 - return LL_SUPER_MAGIC from statfs for the filesystem type (1972)
5649 - updated kernel patches for hp-2.4.20 kernel (2681)
5651 2004-02-07 Cluster File Systems, Inc. <info@clusterfs.com>
5654 - fix truncated write corruption (2366)
5655 - fix for failed assertion in iopen_connect_dentry (1792,2517)
5657 - don't flag the ptlrpcd thread with PF_MEMALLOC (2636)
5658 - ensure len(uuid) < 37 in lmc (1171)
5659 - fix ia64 OOPS in llog_test (2255)
5660 - zero end of page at obdfilter for partial page writes (2648)
5661 - don't leave stale dentries around after renames (bug 2428)
5662 - fix timeouts when evicting a client with a single lock held (2642)
5663 - set deadline for the initial HELLO message to drain (2634)
5664 - print out dotted-quad IP addresses in the socknal (2302)
5666 - additional debugging for MDS client eviction problem (2443)
5667 - fix mkfsoptions support for osts (2603, 2604)
5669 2004-01-27 Cluster File Systems, Inc. <info@clusterfs.com>
5672 - add series for the vanilla 2.6.0 kernel
5673 - add series for the vanilla 2.4.24 kernel
5674 - add series for a cray x86/64 UL kernel drop
5675 - fix xattr patches for the vanilla 2.4.19 series
5677 - generate true UUIDs in lmc (1171)
5678 - have portals stack dumping break in UML (2466)
5679 - avoid bad dchild deref; avoid inum lock w/o creation (2362)
5680 - allocate with _NOFS in ldlm to avoid deadlock (1933)
5681 - wake callback waiting threads on client eviction (2460)
5682 - Add --ptldebug and --subsystem to lmc (1719)
5683 - update assertion to allow safe interrupt allocation
5684 - set rq_no_resend for cancel requests (2432)
5685 - recalculate ptlrpcd timeout after resend (2494)
5686 - call vfs_rmdir when removing pending directories (2368)
5687 - fix renaming a file to itself (2429)
5688 - lmc creates a default one-stripe lov (2454)
5689 - expand procfs space to handle large clusters (2326)
5690 - increase UML stack to avoid overflow
5691 - update lconf's list of debug and subsystem masks
5692 - fix lfs find --obd (2510)
5693 - /proc tunable for disabling filter read caching (2591)
5694 - stop rpm packages from altering slapd.conf (2301)
5695 - disable nagle in the socknal under 0conf (2578)
5696 - choose mds inode size based on stripe count (2572)
5697 - fix kernel-source rpm problems (2516)
5699 - add --disable-doc to avoid pdf generation (2421)
5700 - update documentation, tests, type-os, comments
5701 - avoid format warnings on ia64
5702 - remove the TOE NAL
5703 - tiny code cleanups by removing unused fields
5705 2004-01-07 Cluster File Systems, Inc. <info@clusterfs.com>
5708 - fix obvious semaphore misuse in as-yet-unused setattr path (2348)
5709 - remove the most blatant lies from BUILDING file (2371)
5710 - change default debug level to reasonable production setting
5711 - reduce client side cache size to reduce cache flush time
5712 - reduce max RPCs in flight to avoid unnecessary file fragmentation
5713 - make TCP zerocopy and pinger support enabled by default (2476)
5714 - sync writes completed after process exits caused crashes (2319)
5715 - maintain correct mount count on the MDS (2356)
5716 - backout 1557, because 2316 wasn't really fixed
5717 - better file I/O statistics gathering in /proc
5718 - don't take unnecessary, deadlock-inducing bug in readpage (2383)
5719 - another kernel patch to fix zero-copy TCP function export
5720 - don't take duplicate lock when processing re-sent getattr (2420)
5721 - lctl uses obd_self_export instead of creating new conn (2353)
5722 - MDS/OST recovery case which requires object creation asserted (2425)
5723 - move lfs from /usr/sbin to /usr/bin in packages
5724 - fix race between mds_client_add and mds_client_free (2417)
5725 - use kmalloc instead of slabs in portals (2430)
5726 - don't create duplicate records when a failover MDS is present (2442)
5727 - remove unnecessary mount age check (2332)
5728 - don't remove directory inodes from locks prematurely (2451)
5729 - don't break if MDS service name is the same as hostname (2103)
5730 - fix races in client write RPC generation when cache full (2482)
5732 2003-12-13 Cluster File Systems, Inc. <info@clusterfs.com>
5735 - remove now-unused request->rq_obd (278)
5736 - if an allocation fails, print out how much memory we've used (1933)
5737 - use PORTAL_SLAB_ALLOC for structures, to get GFP_MEMALLOC (1933)
5738 - add the "configurable stack size" patch to most series files (1256)
5739 - ability to write large log records, for 100+ OST configs (2306)
5740 - fix NULL deref when filter_prep fails (2314)
5741 - fix operator precedence error in filter_sync
5742 - dynamic allocation of socknal TX descriptors (2315)
5743 - fix a missed case in the GFP_MEMALLOC patch, can cause deadlock (2310)
5744 - fix gcc 2.96 compilation problem in xattr kernel patch (2294)
5745 - ensure that CWARN messages in Portals always get to the syslog
5746 - __init/__exit are not for prototype decls (ldlm_init/exit)
5747 - x86-64 compile warning fixes
5748 - fix gateway LMC keyword conflict (2318)
5749 - fix MDS lock inversions in getattr/reint paths (1844)
5750 - fix a rare lock re-ordering bug, which caused deadlock (2322)
5751 - fix i_sem/journal inversion in fsfilt_ext3_write_record (2306)
5752 - DLM race condition prevented some lock evictions (2328)
5753 - ENOMEM detection and retry on socknal sends (2230)
5754 - use GFP_NOFS throughout Lustre, to combat ENOMEM (2230)
5755 - move osc_rpcd into ptlrpc, for use in MDC and others (2329)
5756 - protect MDS inode fsdata with stronger locking; fixes assertion (2313)
5757 - better error messages when a client is rejected during recovery (1505)
5758 - avoid cancelling locks which were never granted, after failure (2330)
5759 - fix i_sem/journal inversion in mds_client_add (2333)
5760 - fix truncate/getattr lock cycle deadlock (2334)
5761 - use rpcd to send close; allows resend after timeout, avoid leak (1897)
5762 - fix two rare exit paths which could leak an l_lock() ref (2321)
5763 - fencepost error in MDS/OST orphan recovery (2226)
5764 - make log record alignment 8 bytes (1988)
5765 - lstripe now fails when requested offset > ost_count (2237)
5766 - ensure that all kernel series have a complete list.h (1607)
5767 - fix crashes in special-file operations (2316)
5768 - lctl create/brw OID mismatch, caused by obsolete filter loop (2339)
5770 - allow configurable automake binary, for testing new versions
5771 - small update to the lfs documentation
5773 2003-12-03 Cluster File Systems, Inc. <info@clusterfs.com>
5775 * fix negative export reference count in fsfilt_sync (2312)
5777 2003-12-01 Cluster File Systems, Inc. <info@clusterfs.com>
5778 * release candidate 0.9.1
5780 - orphans are moved into the PENDING directory for possible recovery
5781 - replayed opens now open by fid for orphan/rename safety (1042)
5782 - last close of an orphan inode generates a transno (683)
5783 - chdir() and mount() now pin the directory entry (1020)
5784 - avoid CERROR in normal ll_setattr_raw() error case (1500)
5785 - discard very old requests without processing them (1502)
5786 - remove some common, well-understood CERRORs (1505)
5787 - require O_DIRECT I/O to be page-sized to workaround IA64 crash (1609)
5788 - clear "grant" flags in OST replies until OST grant code lands (1644)
5789 - fix read performance by not clobbering i_blksize on client (1598)
5790 - fix __ldlm_handle2lock oops by not dereferencing lock after PUT (1625)
5791 - make LRU size a /proc tunable, clears locks when reduced (707)
5792 - fix some lprocfs rot that prevented ptlbd from loading (1732)
5793 - server locks take references on exports now (1558)
5794 - build fixes for 2.4.20-rh trees (1663)
5795 - return an error from lov_create if all OSCs are inactive (1751)
5796 - fix import levels when a reconnect happens without a timeout (1597)
5797 - exit early from mds_open if we get a lookup error (1749)
5798 - partial page read at EOF wouldn't wait for disk before sending (1642)
5799 - avoid NULL deref in obdfilter when reading page past EOF (1592)
5800 - avoid LASSERT in ll_intent_lock if server failed very early (1090)
5801 - fix LBUG in ll_it_open_error with rc = -2 (1861)
5802 - write/truncate lock inversion (1639)
5803 - Don't auto-load obdclass, portals modules during cleanup (1495)
5804 - fix timestamps from jumping to "now" (1763)
5805 - extra journal assertions (1648)
5806 - add an extra multiunlink test (1771)
5807 - fix read_record/write_record API (1776)
5808 - fix leak of offset_extent, possible incorrect i_size later (1772)
5809 - fix lasserts in mis-matched transnos during open-unlink testing (1541)
5810 - Debugging for the kqswnal_get_idle_tx problems (1820)
5811 - Allow recovery to be attempted multiple times (1536)
5812 - Write out MDS last_rcvd file after it is first created (1600)
5813 - Fix tx_descriptor leak in failed transmit situations (1827)
5814 - ext3 journaling fixes for assertion failure after IO error (1871)
5815 - class_export_put() on freed export after completion AST error (1896)
5816 - Fix revalidate looping in VFS (1322)
5817 - Don't access a freed export during MDS_REINT timeout (1521)
5818 - Add open-unlink recovery support on the MDS (1673,1764)
5819 - Return an error if no MDS data was read from last_rcvd (1946)
5820 - Fix for lookup "." or ".." crash on error (1932,1931,1935)
5821 - Don't setup a disk device that doesn't match exported UUID (317)
5822 - Reduce bulk RPC timeout to avoid cascading client/OST failures (1845)
5823 - avoid committing NULL handle in force close
5824 - local.sh is now a one-stripe LOV configuration
5825 - POSIX utime.4 -EPERM on FIFO not owned by user (56)
5826 - fix ext3 htree duplicate directory entry corruption (1516)
5827 - POSIX creat.13, fstat.1, open.18, stat.3 new file atime/mtime (2020)
5828 - update to new LOV EA format (2097)
5829 - interoperability for different PAGE_SIZE/wordsize (686,1821,1343,2042)
5831 2003-06-15 Phil Schwan <phil@clusterfs.com>
5834 - imports and exports cleanup too early, need refcounts (349, 879, 1045)
5835 - per-import/export recovery handling (958, 931, 959)
5836 - multiple last-rcvd slots, for serving multiple FSes (949)
5837 - connections are again shared between multiple imp/exports (963, 964)
5838 - "umount -f" would hang if any requests needed to be sent (393, 978)
5839 - avoid pinning large req buffer by copying for queued messages (989)
5840 - add "uuid" to "lctl device" command to help upcalls (991)
5841 - "open" RPCs with transnos would confuse recovery counters (1037)
5842 - do proper endian conversion of all wire messages (288, 340, 891)
5843 - remove OST bulk get LBUGs, fix ost_brw_write cleanup (1126)
5844 - call waiting locks callback from LDLM recovery thread (1127, 1151)
5845 - fix ptlrpc_connection leak in target_handle_connect (1174)
5846 - fix import refcounting bug in OST and MDS cleanup (1134)
5847 - if an invalid-at-open-time OSC returned before close(), LBUG (1150)
5848 - fix very unlikely obd_types race condition (501)
5849 - remove osc_open hack for echo_client (1187)
5850 - we leaked exports/dlmimps for forcibly disconnected clients (1143)
5851 - a failure in read_inode2 leads to deadlock (1139)
5852 - cancel ack-locks as soon as transaction is committed (1072)
5853 - fix major leaks and crashes in the bulk I/O path (937, 1057)
5854 - make sure to commitrw after any preprw to avoid deadlock (1162)
5855 - failing to execute a file in a lustre FS would lock inode (1203)
5856 - small DEBUG_REQ fix to avoid dereferencing a NULL (1227)
5857 - don't ASSERT while cleaning up an incompletely-setup obd (1248)
5858 - obd_uuid2tgt would walk off the end of the list (1255)
5859 - on IA64 the osc would give portals incorrect bulk size (1258)
5860 - fix debug daemon ioctl interface; allows daemon on ia64 (1274)
5861 - fix lock inversion caused by new llite matching code (1282)
5862 - limit the number of dirty pages on a client to 10MB (1286)
5863 - timed out locks were not being corrected cancelled (1289)
5864 - fix O_DIRECT above 4GB on IA-32 (1292)
5865 * major user-visible changes
5866 - fail out/fail over policy now controlled by the upcall (993)
5868 - add OBD_PING to check server availability and failure (954)
5869 - lustre messages are now sent in sending host order (288, 340, 891)
5870 - add eadatalen to MDS getattr reply (340)
5871 - OST read replies may contain second buffer, with per-page status (593)
5873 2003-03-11 Phil Schwan <phil@clusterfs.com>
5876 - LDLM_DEBUG macro fix, for gcc 3.2 (850)
5877 - failed open()s could cause deadlock; fixed (867, 869)
5878 - stop cancelling OST locks when files are closed (481)
5879 - overlapping XID spaces caused network corruption (851, 853)
5880 - fix unsafe fsfilt counter arithmetic; change to atomic_t
5881 - setattr_raw added, to do single-RPC, server-side setattrs
5882 - lmc/lconf syntax change for OST UUIDs
5883 - fix crashy race condition between ptlrpc_free_req and osc_close
5884 - don't use request in mdc_enqueue if we hit a timeout (889)
5885 - don't set the inode i_size for regular files from the MDS (896)
5886 - handle out of order completion AST (842)
5887 - don't LBUG if a lock request times out after receiving AST (913)
5888 - avoid d_rehash race in ll_find_alias by rehashing inside dcache_lock
5889 - if a bad lock AST arrives, send an error instead of dropping entirely
5890 - return 0 from revalidate2 if ll_intent_lock returns -EINTR (912)
5891 - fix leak in bulk IO when only partially completed (899, 900, 926)
5892 - fix O_DIRECT for ia64 (55)
5893 - (almost) eliminate Lustre-kernel-thread effects on load average (722)
5894 - C-z after timeout could hang a process forever; fixed (977)
5896 - client-side I/O cache (678, 924, 929, 941, 970)
5898 - READPAGE and SETATTRs which don't take server-side locks get
5901 2003-02-11 Phil Schwan <phil@clusterfs.com>
5904 - Fix ldlm_lock_match on the MDS to avoid matching remote locks (592)
5905 - Fix fsfilt_extN_readpage() to read a full page of directory
5906 entries, or fake the remainder if PAGE_SIZE != blocksize (500)
5907 - Avoid extra mdc_getattr() in ll_intent_lock when possible (534, 604)
5908 - Fix imbalanced LOV object allocation and out-of-bound access (469)
5909 - Most intent operations were removed, in favour of a new RPC mode
5910 that does a single RPC to the server and bypasses most of the VFS
5911 - All LDLM resource ID arrays were removed in favour of ldlm_res_id
5912 - Aggressively cancel local locks on DLM servers
5913 - mds_reint_unlink sends EA to the client if it's the last nlink.
5914 client uses that EA to unlink OST objects.
5915 - mds_reint_{rename,unlink,link} were rewritten to take ordered locks
5916 - recursive symlinks were fixed (439)
5917 - fixed NULL deref in DEBUG_REQ
5918 - filter_update_lastobjid no longer calls sync, which annoyed extN
5919 - fixed multi-client small-writes to a single file problem (445)
5920 - fixed mtime updates during file writes (607)
5921 - fixed vector writes on obdfilter causing problems when ENOSPC (670)
5922 - fixed bug in obd_brw_read/write() (under guise of testing 367)
5923 - fixed Linux OST size reporting problem (444, 656)
5924 - OST now updates object mtime with writes or setattr (607, 619)
5925 - client verifies file size before zeroing page past EOF (445)
5926 - OST now writes last allocated objid to disk with allocation (108)
5927 - LOV on echo now works (409)
5929 - mds_reint_unlink sends a new buffer, with the EA included. this
5930 buffer is only valid if body->valid & OBD_MD_FLEASIZE, which is only
5931 set if a regular file was being unlinked, and it was the last link
5932 - use PtlGet from the target for bulk writes (315)
5933 - OST now updates object mtime with writes or setattr (607, 619)
5934 - LDLM now has a grant-time callback to revalidate locked items, if
5936 - Many MDS operations were reorganized to combat race conditions
5938 - Merge b_intel branch (updated lprocfs code) - now at /proc/fs/lustre
5939 - configure check to avoid gcc version 2.96 20000731-2.96-98) (606)
5941 2003-01-06 Andreas Dilger <adilger@clusterfs.com>
5944 - Fully reactivate OST imports after reconnection (512, others)
5945 - Make sure client sees our -ENOTCONN from mds_handle (513 - partial)
5946 - More graceful error handling for truncating on dead OST (515)
5947 - Don't error out unless we're actually accessing dead stripes (474)
5948 - Fix garbage sizes when stripes are missing (410)
5949 - LRU counters were broken, causing constant lock purge (433, 432)
5950 - garbage on read from stripes with failed OSTs (441)
5951 - mark OSCs as active before reconnecting during recovery (438)
5952 - lov_enqueue and lov_cancel need to handle inactive OSTs (403)
5953 - lfind did not preserve OST order in output (443)
5954 - symlinks cause hung clients, incorrect data (439)
5955 - stop dereferencing request after dropping refcount (457)
5956 - don't LASSERT(spin_is_locked) on non-SMP (455)
5957 - fixes for many rename() bugs
5958 - fstat didn't correctly synchronize attributes (399)
5959 - server must handle lock cancellation during blocking AST prep (487)
5960 - bulk descriptors were free()d too soon (511)
5961 - fix paths in lconf, which would load incorrect modules (451, 507)
5962 - fix confusing lconf 'host not found' error message (386)
5963 - fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
5964 - fix race condition in mdc_blocking_ast() for inode access (526)
5965 - fix lov_unpackmd() unpacking wrong number of stripes (537)
5966 - fix lov_set_osc_active() marking wrong OSC inactive (440)
5967 - fix bad lstripe lov_unpackmd() assertion (fix layering too) (527)
5968 - fix multiple writes of stripe MD to MDS (358, maybe 519)
5969 - fix lstripe in several ways (kernel side) (527)
5970 - fix request leak in ldlm_cli_enqueue (262)
5971 - incorrect OSC was marked inactive after OST failure
5972 - call mds_fs_cleanup before unmounting filesystem (524)
5973 - fix races between taking ns_lock and ldlm_lock_change_resource
5974 - fix races updating LOV export open file list
5975 - fix lov_enqueue error path, avoid decref-ing bad lock handle
5976 - fix recovery NULL deref in ldlm_cli_cancel_unused
5977 - fix some DLM races by using new hash table for lock handles (419)
5978 - permit the client to specify desired inodes, at replay
5979 - duplicate requests when we queue them for replay reintegration
5980 - fix last_rcvd offset calculation
5981 - sync after each recovered transaction, so we always make progress
5982 - never, not always, ERESTART requests without transnos
5983 - store the lov_desc in the MDS, so we don't depend on getlovinfo to
5985 - skip replay if the MDS says that the client is already connected
5986 - don't check for a recovery-enabled export to match lctl's UUID
5987 - don't INC_USE_COUNT for phantom exports
5988 - don't crash when cleaning up phantom exports (567)
5989 - don't double-finish or set replay data for errored mdc_open requests
5990 - abort requests when they time out, so we don't get old replies
5991 - send/receive replies for AST messages again
5992 - if the client says that it doesn't have the lock, cancel it on the
5994 - if we timeout during I/O, don't try to cancel an in-use lock; instead
5995 mark it as destroyed, it will all work out when decref is called
5996 - fix module use counts (22, 581)
5998 - ASTs now expect a reply (server cancels lock on error reply)
6000 2002-12-02 Andreas Dilger <adilger@clusterfs.com>
6003 - fix many simultaneous client startup (392)
6004 - fix dentry->d_it clobbering
6005 - credentials weren't being shipped for readdir/getattr operations
6006 - remove invalid assertions triggered during some concurrent MD
6008 - proper Lustre versions added (336, 389)
6009 - fix memory leak for create error case (398)
6010 - fix LOV locking bug that would get cli/srv out of sync
6011 - fix echo client over LOV (409)
6012 - fix dbench 2, extN refcount problem (170, 258, 356, 418)
6013 - fix double-O_EXCL intent crash (424)
6014 - avoid sending multiple lock CANCELs (352)
6016 - MDS can do multi-client recovery (modulo bugs in new code)
6018 - many updates, edits, cleanups
6020 2002-11-18 Phil Schwan <phil@clusterfs.com>
6023 - fix null d_it dereference (346)
6024 - fix full OST/dbench hang (333)
6025 - fix permission problem with file removal (286)
6026 - fix removal of OSCs from LOV when they fail
6027 - fix NULL deref during bulk timeout (214)
6028 - fix problems related to multiple filesystems on one MDS (241)
6029 - fixed serious subtle metadata locking bugs
6030 - free locks on clients when inodes are removed due to memory
6032 - fix inode pointer in lock data (285)
6033 - partial support for multiple MDS on a single host (241)
6034 - data locks weren't cancelled at clear_inode time (290, 311)
6035 - intent locks could lead to unbounded lock growth (205)
6036 - added a maximum lock count, an LRU list, and a flusher
6037 - fix multiple rename (365)
6038 - properly abstracted the echo client
6039 - OSC locked 1 byte too many; fixed
6040 - rewrote brw callback code:
6041 - fixed recovery bugs related to LOVs (306)
6042 - fixed too-many-pages-in-one-write crash (191)
6043 - fixed (again) crash in sync_io_timeout (214)
6044 - probably fixed callback-related race (385)
6046 - Add capability to MDS protocol
6047 - LDLM cancellations and callbacks on different portals
6049 2002-10-28 Andreas Dilger <adilger@clusterfs.com>
6052 - limit client IOV size to PTL_MD_MAX_IOV (611336, 191)
6053 - defer open object destruction to close time (601981, 138)
6054 - open/close OST file handle in obdo (OBD_MD_FLHANDLE) (601981, 138)
6055 - move LDLM_ENQUEUE/CONVERT back to MDS portal (625069)
6056 - abstract ll_lookup2, fix ll_revalidate2 to use abstraction (256)
6057 - don't call obd_setattr in ll_file_release for destroyed objects
6058 * protocol change to lustre_msg: move |version| and add |flags|
6059 * protocol change to osc_punch: "start" in "o_size", "end" in "o_blocks"
6060 * lock replay: for LDLM_FL_REPLAY trust client to do right thing
6061 * added replay of create, unlink, link and rename operations during
6062 MDS failover; recovery should be much more robust now
6063 * remove failed OSCs from LOVs (only lov_create uses this so far)
6064 * the lustre-HOWTO was brought (more) up to date (582544)
6066 2002-10-23 Phil Schwan <phil@clusterfs.com>
6069 - in-use dentries weren't being reused properly (617851)
6070 - prevent multiple LDLM setup (599178)
6071 - fix LOV size calculations for truncate (617853)
6072 - fix client handling of MDS intent errors (POSIX)
6073 - fix permission bug in lovstripe.c test (624321)
6074 - fix MDS thread deadlock - move LDLM handler to DLM portal (625069)
6075 - truncate past end of file could corrupt data
6076 - proper cleanup after timeouts, crashes, etc (592524, 550815)
6077 - a race in recovery could return ETIMEDOUT to apps (623947)
6078 - building outside the source directory was fixed
6079 * the lustre-HOWTO was brought (more) up to date (582544)
6080 * major progress was made on recovery functionality
6082 2002-10-10 Phil Schwan <phil@clusterfs.com>
6085 - recovery deadlock fix
6086 - rm -rf causes LBUG fix (617817)
6087 - file open by multiple tasks fix (618962)
6088 - directory permissions bugs (602707 and 620007)
6089 - journal_stop fixed with locking (611313)
6090 - O_APPEND failures resolved (618273, perhaps 614459)
6091 - lconf PATH fix (619770)
6092 - IA64 build fix (621450)
6093 - RPC buffer sizes scale with amount of memory
6095 2002-10-01 Phil Schwan <phil@clusterfs.com>
6098 - locks would be cancelled without throwing away data pages,
6099 resulting in inconsistent data (605627)
6100 - inode attributes were not always being refreshed (605627, 612449)
6101 - lconf now continues to cleanup after lctl reports an error
6102 - MDS now enforces user permissions (602707)
6103 - lprocfs cleanup fixed, but not yet enabled (614157)
6104 - fixed infinite server hang, should a client not respond to an AST
6105 - avoid going into recovery if user calls readlink() with a buffer
6106 that's too small (613941)
6107 - AST RPCs no longer require replies (614867) -- this may be changed
6108 - don't crash server if client sends an IOV that's too big (611336)
6109 - fixed lock conversion deadlock (611892)
6110 - fixed the following of symlinks (614622)
6111 * recovery: the server can remove locks from a client that dies, other
6112 clients can make progress
6113 * more extN patch fixes
6114 * compile-time configurable ptlrpc buffer allocations
6116 - collaborative read cache document
6117 - Lustre Lite Performance CDR document-in-progress
6119 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6122 - fix typo in patch-2.4.18
6124 2002-09-20 Andreas Dilger <adilger@clusterfs.com>
6127 - clear ptlrpc request each time in handle_incoming_request()
6128 - unlink of files now destroys the object on the OST
6130 2002-09-19 Peter Braam <braam@clusterfs.com>
6132 * add hard link support
6133 * change obdfile creation method
6134 * kernel patch changed
6136 2002-09-19 Peter Braam <braam@clusterfs.com>
6139 - stack overflow bug in extN fixed
6141 2002-09-18 Andreas Dilger <adilger@clusterfs.com>
6143 * documentation updates
6144 - add man pages for config tools
6145 - update tests/README to describe testing with new config tools
6146 - finish metadata API descriptions
6147 * bug fixes and cleanups
6148 - statfs workaround for 16TB limit
6149 - LOV stripe allocation improved, can stripe on subset of OSTs
6150 - LOV file size/IO offset was wrong for files > 4GB in size
6151 - object EA data was being dropped, caused files to be unreadable
6152 - memory overflow with non-LOV OST caused memory corruption
6153 - fixed regression tests to work with new config tools, obdfilter
6154 - fixed bug when directory size became larger than 1 block
6155 - fixed bug (for single client case) when PWD was deleted
6156 - invalidate local directory pages when doing intent-based ops
6157 - avoid LDLM oops when lock callback contained bad data
6159 2002-09-09 Andreas Dilger <adilger@clusterfs.com>
6161 * documentation updates
6162 * bug fixes and cleanups
6163 - configuration tools
6166 - 64-bit compile warnings
6167 - 64-bit internal statfs data
6169 * test_brw on persistent OST devices
6171 * lprocfs (disabled)
6173 2002-09-04 Andreas Dilger <adilger@clusterfs.com>
6175 * documentation updates
6176 * bug fixes and cleanups
6177 * configuration tools
6179 2002-08-30 Peter J. Braam <braam@clusterfs.com>
6182 * many small fixes to 0_5_4
6183 * io/network handling
6184 * thinkos in MDS operations
6186 2002-08-24 Peter J. Braam <braam@clusterfs.com>
6189 * crucial basic fixes to 0.5.3
6190 * IOR, Iozone work over Elan
6193 2002-08-07 Phil Schwan <phil@clusterfs.com>
6194 * version 0_5_3, our first alpha
6195 * we use the new Portals iovs
6196 * documentation updates
6197 * bug fixes and cleanups
6198 * small changes in the DLM wire protocol
6200 2002-07-25 Peter J. Braam <braam@clusterfs.com>
6201 * version 0_5_1 with some initial stability,
6202 * locking on MD and file I/O.
6203 * documentation updates
6204 * several bug fixes since 0.5.0
6205 * small changes in wire protocol
6207 2002-07-18 Phil Schwan <phil@clusterfs.com>
6209 * delivered as Lustre Light Alpha
6210 * fixed a crash after handling invalid MDS requests
6211 * fixed directory pages for architectures with non-4k pages sizes
6213 2002-07-11 Andreas Dilger <adilger@clusterfs.com>
6214 * release version v0_4_4
6215 * Moves TCP acceptor to be on port 2432 (unused Coda port) instead
6217 * Fixes a number of interruption problems with OST operations.
6218 * Update documentation for portals header changes
6219 * Move all wire protocol structs/defines to lustre_idl.h
6220 * Fixes symlink length bug.
6221 * Add tcpdump to repository.
6223 2002-07-05 Andreas Dilger <adilger@clusterfs.com>
6224 * release version v0_4_3
6225 * Fixes statfs for inodes on extN.
6226 * Fixes bug in runtests which would delete /etc/hosts.
6227 * Use 64-bit object IDs wherever possible (not into VFS though)
6228 Remove ost_get_info, which is unused by lustre, and out of date.
6230 2002-07-03 Peter Braam <braam@clusterfs.com>
6231 * release version v0_4_2 Fixes a lookup error (type not passed)
6232 * move forward to head of Portals
6233 * move forward to latest Lustre kernel
6235 2002-06-25 Peter Braam <braam@clusterfs.com>
6236 * release version v0_4_1. Hopefully stable on single node use.